The network dynamics of co-offending careers

•Co-offenders rarely commit more than one offense together.•Homophily, degree centrality and transitivity increases future joint co-offending.•Crime type specific network structures present differential effects.•Alter's victimization dissuades continued collaboration.•Non-conventional turning points need to be considered in life-course criminology.

Despite the long-standing acknowledgement that crime is a group phenomenon, little research treats co-offending as a dynamic network process. This study analyses the individual and network processes responsible for long-lasting criminal relationships using co-offending dyads from eight years of arrest records in Chicago. Results from proportional hazard models suggest that homophily with respect to age, race, gender, geographic proximity, and gang identity lead to sustained partnerships. Victimization increases the probability of continued co-offending, while the victimization of one's associates dissuade continued collaboration. Supra-dyadic processes (centrality, transitivity) influence the likelihood of continued co-offending. Results are discussed regarding opportunities and turning points.