This blog is a part of our series, "Perspectives in Crime" where we explore leading academic studies that touch on crime data.
More than 200 people were murdered or assaulted with a firearm daily in the United States between 2006 and 2014. With numbers like this, gun violence in America is frequently referred to in media and rhetoric as an epidemic, but what insights and interventions can be afforded by approaching gun violence through the same lens as an actual disease epidemic? A team of researchers from Harvard and Yale set out to answer that question, studying several years of gun violence incidents and arrest records for the city of Chicago, Illinois, USA, and analyzing how gun violence propagates through social networks.
In their study “Modeling Contagion Through Social Networks to Explain and Predict Gunshot Violence in Chicago, 2006 to 2014,” Ben Green, Thibaut Horel, and Andrew Papachristos used crime data to build an understanding of known associates and social networks among offenders in communities predisposed to gun violence. Criminologists and sociologists often use demographic breakdowns regarding race and economic status to speculate on and observe who is at risk for gun violence victimization. What they have shown is that gun violence concentrates in underprivileged urban communities, and young black men are victimized by gun homicide at a rate 10 times as high as their white counterparts.
Green et al. built the social network from the study on co-offenders who had been arrested together for the same offenses over a period of 8 years in Chicago, a city with longstanding gun violence problems that are intensely concentrated in a small number of disadvantaged neighborhoods.
After building this social network, the researchers analyzed shooting data to see if they could track gun violence as a social contagion, similar to how public health officials would trace the spread of a disease.
Much of the research on the mechanisms of gun violence has measured the spatial diffusion of gun violence from neighborhood to neighborhood. This spatial research peripherally discusses interpersonal relationships like gang activity and the role of drug markets in the diffusion of gun violence, but most related statistical models conceptualized violence as something like an airborne pathogen that moves between neighborhoods.
Co-offenders typically share preexisting social ties, as well as overlapping risk behaviors. Green et al. postulated that individuals are exposed to gun violence through social interactions with previous subjects of gun violence; someone who has been shot is more likely to be embedded in networks and environments with a predisposition towards gun violence. Associating with subjects of gun violence and co-engaging in risk behavior (co-offending) increases a probability of being exposed to/victimized by gun violence. In other words, “when someone in your network becomes a subject of gun violence, your risk of becoming a subject of gun violence temporarily increases,” the researchers stated.
Researchers pitted models treating gun violence victimization as a social contagion against models solely relying on demographic breakdowns of violence dispersion. They found that models based on both social contagion and demographics outperformed alternatives. These models based the probability of future shootings on the history of past shootings, noting that seasonal factors alter rates of violence within the network. Green et al. used prior models and epidemiological research as a basis to assume that gun violence is more likely to spread among the network immediately following a shooting.
To test models, the researchers calculated the risk of every individual in the network being shot, for every day of the study period, based on the data leading up to that day. Each model selected the highest-risk individuals, identifying the top 0.1%, 0.5%, and 1% to be shot each day. Green et al. subsequently evaluated the performance of each model against real-life results.
The social contagion model performed well with the network built from crime data. Green et al. found that 63.1% of the 11,123 gunshot violence episodes in the network during the study period could be attributed to social contagion. On average, subjects of gun violence were shot 125 days after a co-offender was involved in gun violence. The social contagion model identified 4,107 cascades of violence and demonstrated how gun violence propagates through a co-offender network.
Figure 1 shows three example cascades of violence in the study period.
As shown in Figure 2, the social contagion model outperformed the demographics model in predicting the top 0.1%, 0.5%, and 1% of probable victims. But the model combining social contagion and demographics outperformed both single mechanism models. The combined model identified 6.5% of subjects of gun violence when selecting the 1% highest risk daily. Across the three daily high-risk population sizes, the combined model identified 71.7%, 65.5%, and 53.3% more victims than the demographics model alone.
This study found support for thinking of gun violence as an epidemic, discovering that the majority share of gun violence happens among a relatively small share of a cities’ population. The victims of this violence are often highly connected to one another, and the violence is often transmitted through networks by social interactions.
The research suggests that a public health approach or broader security approach can be developed to help mitigate gun violence. Violence prevention efforts should analyze the social dynamic of violence, offering public health and law enforcement salient guidance as to where preemptive intervention could be most effective.
Green, B., Horel, T., Papachristos, A. (2017). Modeling Contagion Through Social Networks to Explain and Predict Gunshot Violence in Chicago, 2006 to 2014. JAMA Intern Med.