Investigating co-offender connections from offender’s hometown background: a practical study in Beijing

Abstract

Co-offending is a common phenomenon in criminology. In previous years, countless studies have contributed to describing the co-offending cooperator's selection process. It has been established that personal characteristics such as age, sex, social/economic status, social/spatial distance, and so on play important roles in co-offender group formation; however, the geographic background of members of the co-offending groups has not received much attention. This study analyzes the assumption that the offender's hometown's homophily enhances co-offending generation and that the size of the offender population of the same hometown (OPSH) would positively impact co-offending formation. To test this, two types of co-offending connections, cross-hometown co-offending (CHCO) and homophily-hometown co-offending (HHCO), are defined respectively, and residential burglary reporting data from Beijing Police from 2005 to 2014 are utilized to investigate the relationship between co-offenders from their hometown backgrounds. The results indicate that HHCO cooperation is more likely to be generated for those who come from large OPSHs, while conversely a large proportion of CHCO cooperation is generated for smaller OPSHs. Additionally, large OPSHs have diverse co-offending connections and attract more partners from the smaller OPSHs, suggesting their more centralized role in co-offending networks.

Appendix

• (1) The definition of Direct centrality is:

  $${C}_{d}\left(u\right)=\sum_{v=1}^{n}r(v,u)$$

  (A1)

  where r(v,u) is a binary variable indicating whether there is the connection between nodes u and v; n represents the total number of nodes in the network. The larger the degree centrality of a given node being, the more neighbors the node has in the network, therefore, the node plays more central role in the network and has more impacting.

• (2) The Betweenness centrality of node i in a network G is:

  $${C}_{b}\left(i\right)={\sum }_{s,t}\frac{{n}_{s,t}^{i}}{{n}_{s,t}}$$

  (A2)

  where s and t represent any pair of nodes in the network except node i; n_s,t is the total number of geodesics from s to t; and nⁱ_s,t indicates the number of the shortest paths between nodes s and t which pass through node i. The greater the BC of a given node, the more it can play the role of information intermediary and plays the central role in the network.

• (3) The Closeness centrality of node k in network G is expressed as follows:

  $${C}_{c}\left(k\right)=\frac{n-1}{{\sum }_{j}{d}_{j}}$$

  (A3)

  where d_j is the distance between nodes j and k, and n is the total number of the nodes in the network G. The greater the CC of a node, the closer the node is to other nodes and the easier it is to transmit network information and resource, so it is more likely to be in the center of the network.

• (4) The formulation of network assortativity is:

  $${f}_{nn,i}=\frac{1}{{C}_{i}}\sum_{j\epsilon u(i)}{f}_{j}$$

  (A4)

  where f_nn,i is the average feature value of all adjacent nodes of node i; u(i) represents the set of adjacent nodes of node i; c_i represent the degree of node i. The correlation level between the feature f_i of each node and f_nn,i reflects assortativity of the network, which reflects the connection preference among nodes.