202 | Andrew Papachristos on the Network Theory of Gun Violence
What this covers
Andrew Papachristos, a sociologist at Northwestern University, presents a network-science framework for understanding urban gun violence. The episode centers on his argument that shootings are not randomly distributed but cluster densely within social networks—in Boston, 85% of fatal and non-fatal shootings involved fewer than 700 people, less than 4% of the city's population. Rather than viewing such violence as either random street crime or the product of broad structural factors alone, Papachristos proposes that network analysis can identify individuals at imminent risk through their proximity to others who have been shot, enabling public-health-style interventions analogous to contact tracing. The discussion moves from epidemiological patterns to practical implementation, examining how the same analytical tools can be deployed by police, hospitals, and community organizations with radically different outcomes.
The conversation spans several distinct domains. Papachristos clarifies that most gun homicides involve people who know each other, not strangers, and explains how violence spreads through networks like a blood-borne pathogen—relational rather than random. He distinguishes assaultive violence between acquaintances from mass shootings and intimate-partner killings, each requiring separate responses. He addresses why the United States suffers gun homicide rates tens of times higher than peer nations despite similar rates of robbery and assault, pointing to gun availability and self-protection culture. The episode examines Chicago's persistently higher homicide rate as a product of gang history, gun sourcing from nearby cash-and-carry states, and uneven political machine allocation of services. Papachristos emphasizes that identifying who faces immediate danger requires combining network maps with local knowledge—outreach workers, teachers, and community figures who understand whether circumstances have changed—and argues that effective intervention requires simultaneous action on multiple fronts: police reform, trauma response, community prevention, and structural change.
Papachristos argues that urban gun violence is highly non-random and concentrates in very small social networks, so identifying who is at imminent risk through network analysis enables focused, public-health-style triage that can save lives while broader structural causes are addressed.
- The majority of gun homicides are assaultive incidents between people who know each other, not random stranger violence
- In studied cities, the large majority of shootings occur within networks comprising under 4% of the community population
- Your probability of victimization is directly tied to how many people in your network are being shot, creating predictable cascades
Gun violence concentrates in small, traceable social networks where shootings cascade through relational ties, not randomly.
- Across many cities studied, gun violence concentrates in very small social networks: in Boston, 85% of fatal and non-fatal shootings occurred within a network of 700 people (under 4% of the community), and in Chicago around 75% of shootings occurred within a network comprising less than 4% of the population.
“in our study of Boston, 85% of all fatal and non-fatal shootings happened in the network of 700 people, less than 4% of this community's population.”
- The networks studied are constructed from instances of co-arrest or being stopped together by police—observations made by police at a point in time and space, setting aside culpability—where each shared event creates a tie, stringing together a network of risky behaviors, while accounting for biases in police data such as officer-initiated 'failure to comply' charges versus civilian-reported crimes.
“The networks are created through instances of co-arrest or being stopped together by the police, so it is actually setting aside anything around culpability, it's an observation made by police at a point in time and space.”
- Because the networks are built only from arrest records and most incidents are never reported to police, the network estimates severely underestimate the true relationships—people who commit crimes together are usually kin, neighbors, or long-known associates—making the analysis conservative relative to richer qualitative data.
“our estimates are extremely conservative if you have information like you're talking about”
- Gun violence spreads relationally, more like a blood-borne pathogen than an airborne one such as COVID, so it follows traceable network patterns and is more containable than airborne contagion, which is what street outreach workers exploit by placing themselves within networks to interrupt disputes.
“Gun violence is relational. So it does not... It's more like a blood-borne pathogen than an airborne pathogen, but you're still tracing these patterns.”
- An individual's probability of being a gun violence victim is directly related to how many people around them in their network are getting shot, and this produces cascades or chain reactions where one shooting leads to subsequent shootings days or months later.
“your probability of being a victim is directly related to how many people around you are getting shot.”
- Stranger homicide is rare; about two-thirds of gun killings are committed by people known to the victim and the remaining third are simply unsolved, meaning most violence occurs within social networks and small communities and is therefore tragic but not random.
“Two-thirds that we know of are by people that are known to the person who is injured or killed. And the other third, we actually just don't know.”
- In violence networks most people have only one or two ties and age out of crime around 26-27, while a small number of high-degree hubs or 'super-spreaders' connect disconnected parts of the network; these hubs are especially vulnerable because moving among multiple rival crews exposes them to greater risk, but they are not kingpins—most people involved are just regular folks.
“Most people have one or two ties. And then there are some people that have a lot of ties. And the people that have a lot of ties are hubs... We can think about super spreaders”
America's gun homicide epidemic stems from more guns than people and a self-protection culture, despite evidence guns increase personal risk.
- The United States is not unusual relative to peer countries in robbery or non-gun assaults, but its gun homicide and gun assault rates are tens of times higher than peer nations, driven partly by having more guns than people and a culture of arming for self-protection.
“We are unique when it comes to our gun homicide and gun assault levels of violence, which are tens of times higher than some of our other peers”
- Although the number one reason people carry or own a firearm in the US is for protection, owning and carrying guns actually makes you less safe, yet the belief that guns keep us safe persists regardless of where people live.
“the tricky science part is actually owning and carrying guns actually makes you less safe”
- People join gangs for the same reason they buy guns—protection and safety in numbers—but because gang members are more likely to have access to weapons, this creates a dangerous mix where the more involved a person is, the more likely they are to be harmed, revealing how deeply unsafe people feel.
“the fundamental reason people join gangs, is for protection, which is the same reason people buy guns.”
Reducing gun violence requires simultaneous intervention across police reform, community prevention, trauma response, and structural conditions.
- Research shows that policing associated with gun violence reduction is limited, small and precise—focusing on a small number of people, places, and behaviors—which is the opposite of broad stop-and-frisk; focused policing is more impactful than expansion of policing powers, though it carries risk if the wrong people or behaviors are targeted.
“policing that is known to have an impact on gun violence tends to focus on a small number of people, a small number of places and a small number of behaviors, which is an argument for more focused rather than broadened policing powers”
- Modern Chicago street gangs are mostly small, informal, neighborhood-based crews—roughly 900 of them, with perhaps only a handful actively involved in organized high-level crime—a departure from the large, organized, politicized gangs of the 1990s, and the policy danger is treating the 890+ small crews as if they were the few organized ones.
“There are, by some estimates, 900 small cliques and crews in the city of Chicago, and most of them, again, are bound to a particular neighborhood”
- Chicago is special and the greatest city in the world.
“Chicago is special. Let's just be clear, it's the greatest city in the world.”
- The summer 2020 increase in gun violence occurred across the entire US, with some of the highest increases in cities like Buffalo, Flint, and New York; Chicago had high absolute numbers but is usually not the highest per capita.
“Some of the highest increases happened in cities we're not talking about Buffalo, Flint, New York.”
- When a shooting occurs, saturating that part of the network with a trauma response can be life-saving by mitigating retaliation (which happens in about a third of Chicago cases) and reducing the trauma of those involved, and hospital-based programs that treat victims' and families' trauma immediately and connect them to services are among the most promising interventions.
“When there's an event, when there's a shooting, saturating that part of the network with a trauma response could be life-saving, which is you can help mitigate retaliation, which happens in about a third of the cases in Chicago”
- Reducing gun violence requires doing many interventions at once—police reform plus investment in community-based violence prevention plus mitigating structural conditions like segregation—because stopping bleeding within a network without addressing entrances into the network is like trying to stop water coming through a dam.
“you're just gonna try to stop the water from coming through the dam in some ways.”
Gun violence takes distinct forms requiring different responses: suicide dominates lethally, while homicides split among mass, intimate partner, and network-embedded assault.
- Gun homicides fall into at least three distinct types requiring different responses: mass/spree shootings, intimate partner violence, and assaultive violence between people who know each other, with the last being the lion's share and most publicly visible.
“the lion's share of gun homicides, which is essentially assaultive, assaults. People that know each other, arguments, disputes”
- Suicide is the most common form of lethal gun violence in the United States, occurring several times more frequently than gun homicides.
“suicide, which is the most common form of lethal gun violence in this country”
- Mass shootings are hard to understand through the same clustering lens because in most cases the target has a network-related purpose—a former or current school, a lover, a former workplace—so it is not a random connection, but it is not as clustered as street violence networks.
“in most cases, the target has some network-related purpose, my former school, my current school, a lover, a workplace that I was involved in.”
- The network logic does not transfer directly to suicide, which is almost the inverse: among high-school-aged students, those with sparse or no networks have more suicide attempts, unlike the dense-tie clustering seen in interpersonal gun violence.
“we actually know that those that have sparse networks or no networks have more suicide attempts.”
Handgun diversion from legal to illegal circulation through theft and unregulated sales fuels persistent interpersonal gun violence.
- Interpersonal gun violence is associated specifically with handguns that are diverted from legal to illegal sources through theft, trade, or unregulated private sales, and because guns are durable goods they persist and circulate, unlike consumable goods such as drugs.
“Guns are by and large durable goods, they're not like, say, drugs or some other food, whether you eat them, you ingest them, they're gone, you get more. Guns stay around for a long time, so they get... They move around.”
- Chicago's persistently higher homicide rate is driven by a combination of factors beyond segregation and racist housing policy: a uniquely deep and organized history of gangs, more recovered guns due to proximity to cash-and-carry states like Indiana, the continued dominance of a political machine system that allocates services unevenly, and a lack of police reform compared to cities like Los Angeles.
“What keeps Chicago unique, one of the things outside of segregation, racist housing policies... is its history of gangs in the city, which is fundamentally different than some of those other cities.”
- During the June 18-July 5, 2020 surge in Chicago (416 shootings, 74 fatal), violence did not appear in new places but reactivated old, well-worn network pathways among already high-risk groups and individuals, with COVID-related changes in routine activities and increased gun carrying as contributing factors.
“all of these old well-worn pathways, these old networks that had been in the neighborhood, sometimes active, sometimes un-active, started popping up.”
Network analysis becomes actionable only when combined with local expert context about relationships, employment, and changed circumstances.
- Network maps only become actionable when combined with local experts—outreach workers, teachers, preachers, social workers, cops—who supply the context behind ties (history, relationships, whether someone already has a job or has turned their life around), unlike simply 'working a list' which ignores whether a person's circumstances have changed.
“the context becomes important, and you only get that by engaging local experts, good preachers, teachers, cops, outreach workers, social workers... That's what makes this information actionable”
- Traditional risk-factor approaches (being young, poor, Black or Latino, male, living in a disinvested neighborhood) describe enormous populations—up to 90,000 people on Chicago's west side—yet only a few hundred actually get shot, so structural conditions alone cannot identify who is in harm's way today; network analysis bridges from structural risk to specific actionable identification.
“those risk factors describe 90,000 people... only a few hundred will get shot”
- The same network science can be deployed very differently by police, public health agencies, and community-based entities; police use has gone poorly because they 'work a list' and treat victims like offenders raising constitutional and moral concerns, whereas in public health the science is essentially contact tracing—finding who is in harm's way and getting them resources—as done successfully with HIV and other epidemics.
“In the public health context, the science is exactly the science of contact tracing, it's literally finding out who's in harm's way, getting resources to keep them safe”
Gun violence victims experience trauma distinct from war veterans because they return to the same environment where violence occurred and among perpetrators.
- Unlike combat veterans who return from a war zone to a theoretically safe home, gun violence victims go right back to the same block where they were shot, re-traumatized by living among the people who shot them and by repeated nearby shootings, making this trauma context distinct and requiring dedicated investment.
“A lot of the people who are caught up in shootings, they go right back to the same block, they were shot 600 feet away from their home and they walk past that every day.”
- Most gang criminal activity is cafeteria-style and haphazard rather than well-planned; members spend most of their time in compliance with the law doing ordinary things, but violence is ever-present, sporadic yet traumatic, consuming their energy and re-traumatizing them like a ghost on the block.
“the best way to describe delinquent or criminal activity of gangs is a cafeteria style, which is most groups do a little of this, a little of that.”
- Many supposed 'stray bullet' victims, including children, were not random targets but were standing next to the intended target—such as a seven-year-old shot while holding the hand of a 27-year-old who was the real target—reinforcing that even tragic bystander deaths are non-random and network-related.
“many of these sort of stray bullet incidents aren't that far stray, they're tragic. They weren't the intended target, but they were literally standing next to somebody who was”
- Crime rates are far more than individuals making bad decisions; while people do make real choices (carrying a gun when you feel unsafe is a reasonable choice), the deeper questions—why there are so many guns in the neighborhood, why the state fails to protect, why one is in harm's way—are contextual, so policy that focuses only on getting individuals to make better decisions is overly simplistic and a program can 'fail' even when an individual changes their behavior if the surrounding network does not.
“crime rates are way more than people making bad decisions. That's an overly simplistic view of crime and violence.”
Social media accelerates disputes through constant public visibility and broad reach, though the causal link to offline violence remains unclear.
- Social media is used to taunt and disrespect rivals, honor friends, and express grief, and while the vast majority of online conflict never becomes physical violence, it appears to make disputes ramp up more quickly because it is non-stop, public, and reaches a broad audience, though the online-to-offline causal link is hard to establish.
“The vast majority of these things that happen online never end up in violence.”
- Concern that new media drives youth violence is a recurring historical pattern—from 1940s senatorial hearings on comic books and MAD Magazine to hearings on Ozzy Osbourne and NWA—and current alarm over social media is the latest iteration, distinguished mainly by the immediacy and speed of social media.
“The 1940s, there were entire senatorial hearings on Mickey Mouse and comic books and MAD Magazine, because it was a new form of media”