May 21, 2010
A sharp increase in Pittsburgh homicides in 2003 led local leaders to launch an antiviolence initiative, One Vision One Life. RAND Corporation and Michigan State University researchers used available data on homicides, aggravated assaults, and gun assaults to test program effects in three neighborhoods. They found no program effects on homicide and negative effects on assaults. Possible explanations for these outcomes include improper program implementation, failure to address gang structures, and data limitations. The findings raise questions for initiatives and evaluations elsewhere.
In 2003, a record 125 homicides in Allegheny County, Pennsylvania, including 70 in the city of Pittsburgh, raised concerns among community leaders regarding the level of violence in the area. In response, local leaders created the Allegheny County Violence Prevention Initiative, later called One Vision One Life.
One Vision was similar to violence-prevention programs in other cities, including Boston, Chicago, and Baltimore. Such initiatives use street-level work and intelligence to become aware of and intervene in potentially violent disputes. One Vision staff members met with those on a Chicago project to learn how best to implement the program.
Programs such as One Vision seek to address the violent "code of the street" prevalent among many inner-city youth, a code developed in response to a lack of legitimate, successful role models. One Vision's focus was a six-point plan to stop local shootings. The plan included
One Vision work is performed by an executive director, a program director, five area managers, and more than 40 community coordinators, and supported by a data manager. Most staff members were raised in and therefore are intimate with inner-city street life and the "code of the street."
RAND Corporation and Michigan State University researchers assessed the effects of the program in three areas of Pittsburgh: Northside, the Hill District, and Southside. All three have per capita incomes below the national average, and two of the three have homicide rates above the city average. Northside is the largest of the three and features a critical hub of legal and illegal activities in the city. It is also undergoing gentrification, a process leading to some community conflict. The Hill District, once a thriving, prosperous, and influential black neighborhood, has suffered a precipitous decline and now has issues with guns, drugs, and individual or group disputes. Southside does not have a homicide rate as high as those in Northside or the Hill District, but its geography and topography help shelter many illegal activities, including drug dealing.
Within the target communities, community coordinators worked with clients who were typically male, black, about 18 years old, and in need of a wide variety of assistance and services. Fifty percent did not have a job, and 30 percent had a substance-abuse problem. Yet, most were not at high risk for violence, having not been violent recently, in a gang, or in the criminal-justice system. In response to their perceptions of community risk for violence, community coordinators would undertake actions ranging from conflict mediation to outreach to community rallies against violence.
To measure the effect of the program on local violence, the researchers used a propensity-score analysis enabling them to compare One Vision neighborhoods with others in the city. They also compared the effects of the program with those suggested by One Vision staff members as being most similar to the analysis areas. Finally, they assessed any "spillover" effects of the program either displacing violence or extending crime-suppression benefits from the target communities to surrounding areas. (Because Northside is largely isolated within the city by the Ohio and Allegheny Rivers, the researchers did not test for spillover effects there.)
The researchers measured changes in homicide, aggravated assaults, and gun assaults before and after the intervention, controlling for neighborhood attributes, seasonal effects, and trends over time. Rather than finding that One Vision was associated with a measurable reduction in violence, they found the program to have no effect on homicide rates and to be associated with increases in aggravated assaults and gun assaults in all three areas. They also found some "spillover" effects in surrounding areas; specifically, aggravated assaults increased in the "spillover" area neighboring Southside but decreased in that neighboring the Hill District, while gun assaults increased in both the Southside and Hill District spillover areas.
It is difficult to explain why a program did not have desired effects, much less effects opposite of those intended. Nevertheless, the researchers offer some explanations for why the program may not have had its intended effect, or at least why there is no quantitative evidence of its intended effect.
First, all evaluations of this sort face difficulties in identifying best comparison areas, measuring program delivery and performance, and isolating program effects. Truly random design and analysis is generally not possible for such assessments. Quasi-experimental design can approach the rigor of random design but cannot control for all variables that might affect levels of violence.
Second, the implementation of One Vision deviated in several ways from ideal implementation. One Vision lacked consistent documentation, and its staff rarely used the available documentation in any systematic way to guide program actions. Community coordinators focused more on persons in need of services than those at risk of violence. This and the difficulties in program and evaluation design may be related. One Vision, by providing youth programming, may have some long-term success in helping youths avoid violence. The program design and evaluation, however, were focused on a more immediate reduction of violence, a reduction the researchers did not observe.
Third, the program did not do much to address any group or gang structure generating violence. The Chicago program on which the Pittsburgh program was partially modeled explicitly focused on gangs. The Boston program and similar programs in several other cities had a group accountability component that was lacking in the Pittsburgh program. It may be the case that the gang structure in Pittsburgh is more fluid and would require a different approach from that used in Chicago. It also may be the case that difficulties in comparing the degree to which individuals are at risk for violent behavior in these cities limits analyses such as this.
Still other study limitations may have affected these findings. While One Vision's focus is on reducing homicides and shootings, only direct measures of homicide were available. Homicide is a rare occurrence; detecting measurable changes in it is therefore difficult, as would be measuring gang-related and non–gang-related incidents. The available control measures, based on 2000 census data, could not measure demographic and socioeconomic changes that have since occurred in these neighborhoods.
These findings raise several critical issues for similar and future initiatives. Among others, these include the transferability of success in programs elsewhere and elements missing in the Pittsburgh implementation. Successful results from Chicago and Baltimore programs suggest the promise of these programs, while the Pittsburgh results suggest the need for continued rigorous evaluation.