Examining the Differences between Police-Defined and Network-Defined Co-Offending Groups

Alexandra Ciomek

Contact: ciomek@fas.harvard.edu

Law enforcement agencies are pivotal institutions in urban areas beyond their influence on the criminal justice system. They can also both shape and define the social networks of citizens, especially for those that partake in criminal activity. Using law enforcement data from Boston, MA, I extract the co-offending network of individuals with contact with law enforcement from 2007-2014 and use community detection techniques to obtain “network-defined” co-offending groups. I compare these groups to police-defined gangs to identify the differences between the two classifications and show how the different boundaries of the groups affect predictions of individuals’ offending patterns. It is well established that peers can affect an individual’s delinquent or criminal behavior. Furthermore, gang membership influences delinquency beyond the effects of associating with delinquent peers. We therefore expect gang membership and co-offending patterns to be predictive of individual offending. In practice, official records of gang membership are typically based on law enforcement assessment. But do police-defined groupings actually improve predictions of individual offending? Or do network informed groupings perform as better predictors? To answer these questions, I analyze a co-offending network of individuals involved in arrests and Field Interrogation and Observation contacts with the Boston Police Department. The findings show that some gang members are also in network-based offending groups, but also suggest that current gang classifications may not capture all criminal groups, at least in the context of co-offending. Given the impact of criminal justice system involvement on the life course, as well as the added interest from law enforcement that comes with being a gang member, we must understand how current policing practices capture the true nature of offending at the individual level.

← Schedule