Combining Community Monitor Data with Archival Crime Data to Predict Neighborhood Increases in Youth Violence

Introduction:  Community Violence Prevention Interventions such as Chicago’s CeaseFire Project rely on outreach workers and violence interrupters drawn from the neighborhoods they serve, who have credibility with youth likely to become involved in violence, and who thoroughly internalize the anti-violence ethos of the program they represent.  Recruiting, training, and deploying these workers in a manner that accurately meets the needs of communities would be greatly enhanced by the ability to predict which neighborhoods are likely to experience increased youth violence in the near future. 

Methods: In the first phase of this study, drawing on methods from the embryonic field of predictive epidemiology, a model was developed to predict neighborhood change in violent crime based on publicly available archival data of minor offenses likely to be committed by youth (Henry, 2013). This model was predicated on the assumption that change in minor offenses (e.g., vandalism, weapons violations, and disorderly conduct) reflects changes in the intensity of gang conflict, which tends to escalate and produce local outbreaks of youth violence. 

In the second phase of this study, a community monitoring approach was implemented which involved collecting weekly incident observations via telephone from a fixed panel of geographically dispersed community members over 24 weeks. Each reported incident was geocoded and tallied. Then, all reports were entered into a model in conjunction with police data to improve the accuracy of predicted violence at the census tract level.

Results: The prediction model was able to identify neighborhoods likely to experience an increase in youth violence with a positive predictive value of .71 for increases of one or more violent crime in all neighborhoods, and .65 for increases of one or more violent crimes in neighborhoods that had not had increased violence in the preceding three months.    Incorporating information from the community monitors increased the positive predictive value from .78 to .66 in the Englewood community

Conclusion: The prediction model combined with data from neighborhood monitors shows promise for more accurate use of scarce violence prevention resources and for community-based policing efforts.  The community monitoring system is not expensive to operate, but does require some financial and other resources.  Important next steps are to investigate the optimal group of minor offenses to use as predictors, the optimal time period between predictors and criteria, and the optimal time window for observing change in predictors.