Study 1
Chalfin and colleagues (2017) conducted a randomized controlled trial to assess the effectiveness of Street Lighting in New York City (NYC) Public Housing on complaints of index, felony, and misdemeanor crimes and assaults, homicides, and weapons crimes. Specifically, they tested whether developments that received a greater dosage (number) of lighting experienced larger reductions in crime.
In May 2015, 80 housing developments were identified for inclusion in the study based on elevated crime rates and perceived need. A total of 39 developments were randomized into the treatment group and 38 into the control group through paired random sampling methods. The groups were stratified based on the number of outdoor nighttime crimes in a development in the two years prior to the intervention.
Between February 29 and March 7, 2016, approximately 400 light towers were deployed to outdoor public spaces in New York City Housing Authority (NYCHA) housing developments by using a heat map developed from resident and law enforcement recommendations. Among treated developments, the allocated dosage was measured by the number of square feet per light. A minimum of two lights were allocated to each of 39 treated developments, and the 320 remaining lights were randomly assigned, per square feet of uncovered areas, to the treated developments. Thus, allocated dosage was slightly different for each site. To protect against bias, assigned dosage of lighting for each development was used as the measure of lighting. The control group developments received no additional outdoor lighting.
The study used New York Police Department (NYPD) complaint data from March 2011 to August 2016. Coordinates were plotted, using a statistical software package, to calculate the distance between complaint location and the footprint of the closest development. Complaints that occurred in the treatment and control developments and in the surrounding communities were included in this study. In the main displacement analysis, complaints that were within 750 feet of a development, which did not occur on the development grounds, were also included. The data for each complaint included the type of offense, date, time, and type of place where the incident occurred. Complaints for sex offenses and rapes were not included in this study.
The study focused on four main outcomes: 1) index crime complaints; 2) felony crime complaints; 3) assault, homicide, and weapons crime complaints; and 4) misdemeanor crime complaints. Index crime complaints were defined using an offense type variable and included murder and non-negligent manslaughter, negligent or unclassified homicide, robbery, felony assault, burglary, and grand larceny of a motor vehicle. Felony and misdemeanor complaints were defined using the law code category variable in the NYPD data file. Assault, homicide, and weapons complaints were defined using an offense type variable and included murder and non-negligent manslaughter, negligent or unclassified homicide, assault and related offenses, felony assault, and complaints for dangerous weapons. The assault, homicide, and weapons crime complaints were used as a measure of interpersonal violence.
For each complaint type, four primary crime locations were examined: 1) outdoor nighttime crimes, 2) indoor nighttime crimes, 3) outdoor daytime crimes, and 4) indoor daytime crimes. Outdoor and indoor crimes were defined using a location description variable. Outdoor crimes included complaints that occurred in front of, opposite, or in the rear of a building, while indoor crimes included complaints that were designated as occurring inside a building.
Regression analyses were conducted for this study. Data was examined by year, month, and day. Only post-implementation data, from March 2016 to August 2016, was included in the analysis, but data from 2011 to 2015 was used for control variables in the regression analysis. Development-level characteristic data was drawn from NYCHA to control for population rates, development size (height, density, and square footage), males between the ages of 15 and 24, precinct, housing units per population, number of entrances per building, average household size, and whether the housing development had an elevator or was a walkup. The study authors used NYPD data to control for differences in law enforcement presence. They did not conduct subgroup analyses.