McGarrell, Chermak, Wilson, and Corsaro (2006) analyzed the impact of the Indianapolis (Ind.) Violence Reduction Partnership (IVRP) using an interrupted time series quasi-experiment to examine homicide trends in Indianapolis. They used six Midwestern cities as comparison groups: Cincinnati, Ohio; Cleveland, Ohio; Columbus, Ohio; Kansas City, Mo.; Louisville, Ky., and Pittsburgh, Pa. Information for these cities was obtained using the Supplementary Victim-Level Homicide Database. These cities were chosen based on their similarities to Indianapolis in homicide and population rates, as well as their proximity to Indianapolis.
Homicide data was studied from Jan. 1, 1997, to June 30, 2001, with the intervention period occurring in April 1999. The postintervention period was May 1999 to June 2001. McGarrell and colleagues used time-series analyses to compare monthly homicide trends during these periods for Indianapolis and the six comparison cities. Preintervention monthly homicide means were compared to postintervention means to determine the effect of the IVRP intervention on homicide trends.
Corsaro and McGarrell (2009) disaggregated monthly homicides into gang- and non–gang-related homicides. Since the intervention was targeted at gang-affiliated chronic offenders, the researchers believed that the impact of IVRP should be stronger for gang-related violence. Otherwise, the decline in homicide rates could be attributed to external forces.
The data for this study was obtained from information on offenders from the pulling-lever meetings. During the meetings, researchers took detailed notes about homicide incidents and which suspects were involved in gangs. These incident review findings revealed patterns of violence in Indianapolis not available from official data sources, and they showed a high degree of gang-related offenses.
Corsaro and McGarrell used an interrupted time-series analysis to differentiate the impact of the IVRP intervention on gang versus non-gang-related homicides. Monthly events of homicides were coded as gang-related or non-gang-related, based on information from researchers and police files. If at least one victim or suspect was shown to be affiliated with a gang, the homicide was coded as gang-related. They compared percentage changes across homicide types between pre- and postintervention periods to analyze the differential effects of the intervention. Monthly homicide trends were assessed between January 1997 and June 2001, with April 1999 as the intervention date.
Corsaro and McGarrell (2010) used a time-series analysis to examine changes in homicide rates between different neighborhoods in Indianapolis, as well as different age, race, and gender groups. The purpose was to isolate the effects of IVRP on certain high-risk groups and to control for certain factors that could have influenced its effect.
Three urban neighborhoods designated as police beats were used to analyze homicide rates in high-risk areas relative to other neighborhoods in Indianapolis. These areas were adjacent to the locations of the pulling-levers meetings, and were chosen as target areas based on several risk factors, including male divorce rate; residential stability; and concentrated disadvantage (which took into account unemployment, poverty, and female-headed households).
City-level neighborhood data was used to examine changes in homicide patterns. Population and demographic information was obtained from the 2000 U.S. Census. All homicides that occurred between January 1997 and December 2000 were included. 1997 and 1998 were treated as the preintervention period, with 1999 and 2000 as the postintervention period. Annual homicide events were studied for each year, and risk-specific homicide patterns were examined before and after the IVRP intervention for each group. Age, gender, and race were studied for each homicide incident, where each incident represented a unique homicide victim. African American male homicides were the total number of annual events that involved African American males aged 15 to 24 as either the victim or suspect in a homicide incident, and this criteria applied to all groups in the 15- to 24-year-old age range. The comparison group was all other homicides, which included all homicides that did not involve 15- to 24-year-olds.