That is a new paper from Andrew V. Papachristos and Christopher Wildeman, here is the abstract:
This study estimates the association of an individual’s position in a social network with their risk of homicide victimization across a high crime African American community in Chicago. Data are drawn from five years of arrest and victimization incidents from the Chicago Police Department. Results indicate that the risk of homicide is highly concentrated within the study community: 41 percent of all gun homicides in the study community occurred within a social network containing less than 4 percent of the neighborhood’s population. Logistic regression models demonstrate that network-level indicators reduce the association between individual-level risk factors and the risk of homicide victimization, as well as improve overall prediction of individual victimization. In particular, social distance to a homicide victim is negatively and strongly associated with individual victimization: each social tie removed from a homicide victim decreases one’s odds of being a homicide victim by approximately 57 percent. Findings suggest that understanding the social networks of offenders can allow researchers to more precisely predict individual homicide victimization within high crime communities.
Some of those sentences could be framed for their importance. For the pointer I thank DP.