This study aims to assess three related aspects of co-offending networks: (1) the characteristics of co-offending groups and the duration of group offending careers, (2) the impact of membership in co-offending groups on total offending and the length of individual offending careers, and (3) the impact of offender arrest (or changes in co-offending group structure) on the offending patterns of connected co-offenders.

Data on sentenced burglary offenders (N = 270) in one county in Pennsylvania from 2001 to 2010 are used to examine the impact of co-offending group membership, as well as the relationship between the changing network structure and the offending patterns of connected co-offenders, within a two-level modeling framework.

Larger groups with more dispersed offending structures offend over the longest span. Additionally, membership in co-offending groups is associated with more total offending and a longer individual offending career. Finally, the arrest of structurally important offenders, compared to more peripheral offenders, is significantly associated with the decreased offending of connected co-offenders.

The removal of a highly central "instigator" or "recruiter" is associated with desistance among connected co-offenders. Future research should examine the mechanisms behind these effects, and why the arrest of co-offending partners is associated with desistance.