Study
Circo, and colleagues (2020) used a quasi-experimental, synthetic control approach to estimate the impact of Detroit Ceasefire on shooting victimization in certain neighborhoods of Detroit, Mich., above and beyond decreases in violence experienced in the rest of the city. A synthetic control approach was used to identify the causal influence of the intervention on a series of outcomes. Synthetic control designs compare changes in the outcome variables between treated and untreated groups. The study authors used a synthetic control method called “microsynth” to evaluate the impact of the Ceasefire intervention. They chose the “microsynth” method because it was designed for studies with many small observational units. In this study, the shooting data were aggregated to the census block level in the two areas of interest (e.g., fifth and ninth precincts).
The main outcome of interest was shooting victimizations. Data for the analysis were provided by the Detroit Police Department and included detailed information (e.g., time, date, location, circumstances, age, race, and sex) about all fatal and nonfatal shooting incidents from 2011 to 2019. All shooting incidents were linked to a census block and police precinct. In total, the data included 9,699 shooting incidents with 11,271 victims. The post-intervention period of the analysis began during the third quarter of 2015 and the study authors examined four endpoints for analysis purposes. Shooting victims were categorized into the following age groups: 15–24, 25–34, 35–44, 45–54, and 55+. The CrimeSolutions review of this study focused on the effects of the intervention on ages 15–24 and 25–34 at the 1-year follow-up period.
The Ceasefire treatment group included 160 census blocks in the fifth and ninth precincts. The comparison group used for the synthetic controls analysis was selected from all remaining census blocks in the city (
n
= 518). Following matching procedures using both time-varying and time-stable characteristics, a synthetic control group was formed with 160 units. No subgroup analysis was conducted.
Study
Circo and colleagues (2019) used a quasi-experimental design to evaluate the impact of Detroit Ceasefire on individual-level outcomes, focusing on individuals who attended call-in meetings that occurred between August 29, 2013, and December 1, 2016. During this time period, 14 meetings occurred, during which a total of 254 individuals involved in gang and street-group violence attended. Meeting attendance ranged from 9 to 26 individuals.
The primary outcome of interest was time to arrest following call-in meetings (measured in number of days) for any arrest, any violent arrest, and any weapons arrest. The follow-up period was 3 years; thus, the minimum number of days that individuals were observed was 395 and the maximum number of days was 1,095.
To evaluate the impact of the Ceasefire call-in meetings on arrests, individuals who attended the meetings were compared with a matched comparison group. A comparison pool of individuals under correctional supervision who were not subject to the call-in meetings but were known as gang- or group-involved individuals, were selected. This resulted in a total of 335 individuals. In terms of demographics, all individuals in the treatment and comparison groups were Black males.
Next, the treatment and comparison groups were matched using inverse-probability weighting on the propensity scores. This was made possible by the criminal history data provided by the Michigan State Police. Post weighting, the average age of individuals in the comparison group was 24.9 years, and the average age of individuals in the treatment group was 25.4 years. On average, comparison group individuals had 0.50 violent arrests, 0.80 property arrests, 0.62 disorder arrests, 0.20 drug arrests, and 0.26 weapons arrests, while the treatment group had 0.60 violent arrests, 0.78 property arrests, 0.83 disorder arrests, 0.53 drug arrests, and 0.37 weapons arrests. This weighting procedure reduced the covariate imbalance between the treatment and comparison groups.
Finally, to evaluate time to rearrest following the call-in meetings, Cox proportional hazard regression models were used to analyze the 3-year survival rates. No subgroup analysis was conducted.
