Network Position and Police Who Shoot

Officers with who bridge otherwise distinct groups of officers in networks of misconduct are more likely to shoot at civilians.

The most common explanations for acts of police violence tend to focus on the individual attributes of officers who use excessive or lethal force as compared to those who do not - we might think about factors such as poor impulse control, authoritarian personality traits, or differences in risk of excessive use of force that depend on officers' demographic background.

But police organizations are inherently networked - by design, officers work within units and often with partners or in teams. And just like other forms of deviance, problem police behaviors can be traced to officers' networks.

In this study we assessed whether officers' position within networks (of officers who are associated by being co-accused of misconduct) can be used to predict officers' subsequent risk of shooting a civilian. Our goal was to determine whether certain network positions are "risky" and in particular if officers in a brokerage position are more likely to shoot.

We build here on the sociological concept of brokerage, specifically we use measures of "betweenness" brokerage to analyze officers' network position within the Chicago Police Department using more than 38,442 complaints filed against police officers between 2000 and 2003.

Brokers occupy important positions “in between” other actors in a network and oftentimes connect otherwise disconnected parts of a structure. For example, an officer who is transferred from one police district to a detail in a different unit or district would have network connections to two different possible pools of officers—those in the previous district and those in the new district. The transferred officer might, thus, occupy a position as a “broker” between officers in these two districts. Examples of brokerage positions are given by node A in the figure above.

We find that it is exactly these types of network positions (high "betweenness" brokerage positions) that predict an increased propensity to shoot at civilians. Even after accounting for where officers' demographic background, their career trajectories, and the activities that officers are involved in at work (including the geographic location they are assigned to), we find that officers who shoot at civilians are often brokers within networks of officers.

“Officers who shoot at civilians are often brokers within networks of officers.”

We find that each standard deviation increase in betweenness centrality predicts roughly 1.17 higher odds of shooting net individual factors. The remaining association is strongly related to organizational-level factors such as "shuffling" a phenomena we define as the re-assignment of officers across districts or other units. Future research might consider whether this is generated by “problem” officers being moved from one unit or district to the next, which may inadvertently make problem officers brokers and, in essence, “spread” aggressive behavior.

You can read the full article here:

Zhao L, Papachristos AV. Network Position and Police Who Shoot. The ANNALS of the American Academy of Political and Social Science. 2020;687(1):89-112.

You can also learn more about this project at We also talk about our work in the following BBC podcast:

The data used here are made publicly available by The Invisible institute (, a journalistic enterprise based on the South Side of Chicago that aims to increase the accountability of public institutions to citizens. We are happy to share our code and replication materials upon request.