At The Crossroads November 23, 2021

“Violence Is Contagious”: A Conversation with Andrew Papachristos

By Greg Berman


“Gun violence is tragic, but, in the majority of cases, is decidedly not random,” Northwestern University sociology professor Andrew Papachristos told the Chicago Sun-Times earlier this year.

Employing both statistical models and qualitative methods, Papachristos has been able to show that a relatively small number of individuals are involved in gun violence within any given community—and that these people tend to be connected to one another by a web of relationships. 

For example, in a study published in the Journal of Urban Health, Papachristos and two coauthors looked at young men with an elevated risk of gunshot victimization in Boston and found a social network of about eight hundred individuals. On average, individuals in the network were less than five handshakes away from the victim of a shooting. Strikingly, the closer someone was to a gunshot victim, the greater the probability that person would later be shot; each network step away from a gunshot victim decreased a person’s odds of getting shot by approximately 25 percent.

Unlike many academics, Papachristos is committed to translating his research into policy and practice. To facilitate this, he recently helped launch the Northwestern Neighborhood & Network Initiative, which seeks to leverage the university’s expertise to address problems facing the residents of Chicago and surrounding communities. 

In September, Greg Berman, The Harry Frank Guggenheim Foundation Distinguished Fellow of Practice, talked to Papachristos about his research into neighborhood violence and about the challenges faced by academics who choose to venture beyond the ivory tower. The following is an edited transcript of their conversation. 

Greg Berman: A lot of your work, some of which has been supported by The Harry Frank Guggenheim Foundation in the past, has been devoted to network science. I wonder if you might start by explaining what you mean when you use the expression “social network” and in particular what you have found about how homicides tend to cluster in a place like Chicago.

Andy Papachristos: When I talk about network science or social networks, what I’m actually talking about are the social relationships that link people, places, and institutions. There’s a whole field that uses statistical models, as well as qualitative data, to understand how patterns of relationships affect what we do. 

A lot of my work has applied this idea to understanding patterns of crime and violence, specifically gun violence. One of the most robust criminological findings is that delinquency and crime are group phenomena. The same is true of violence.

I started out looking at conflicts between groups and gangs and how stable these conflicts are. This was actually the original work that was supported by the Guggenheim Foundation. It was really looking at how these patterns endure. A lot of the murders that we see today, especially those that involve gangs or groups or crews, are actually determined long before any particular member even joins the group. These structures exist in unseen ways and they actually shape who your enemies are and who your allies are. And so I used network science to figure out if we can understand how these structures essentially inform or predict subsequent acts of violence.

We went from there to looking at individuals to see if we could figure out who’s going to get shot as an individual, not just as a bucket of risk factors. Criminologists and sociologists know a lot about the risk factors associated with violence—being poor, being young, being Black or Latino, living in a particular neighborhood. But when you look on the ground on any given day, those risk factors only take you so far because everybody in a particular neighborhood has risk. So how do we figure out which one or two or three people are going to get shot?

Sometimes people are saying that they feel so unsafe that they have to protect themselves, even though they know carrying a gun is going to potentially get them in trouble, or get them killed. I do think people have agency, but I think those are forced choices in some ways. 

So we went back to try to understand the shape of people’s social networks, and their placement within them, and how that affected their probability of getting shot. The methods that we used borrowed from epidemiology, from the study of infectious disease. When you apply these ideas to violent behavior, homicide becomes an interaction. 

We found that gun violence concentrates within social networks. So a small proportion of individuals are at the center of gun violence within any given community—and by small I mean a couple of hundred people in a community of tens of thousands. Exposure matters: when people around you are getting shot, your probability of being shot skyrockets. Violence is contagious in a very real sense—it cascades through networks in very predictable ways. It actually does spread like pathogens. There are others, but those are the key findings that we’re seeing in multiple cities and that we’re trying now to leverage for violence-prevention efforts.

Not long ago, you gave a talk to a class at Princeton University that was entitled Society Didn’t Do It; Networks Did. I don’t mean to put too much weight on a cheeky title that maybe you didn’t even come up with, but it made me wonder how you think about individual agency when it comes to violence. I have always thought that there was an implicit moral argument in saying that violence is like a disease, since we tend not to hold individuals accountable for getting an infectious disease in the same way that we hold people accountable for shooting someone.

I did not come up with that cheeky title, [Princeton sociology professor] Fred Wherry did. However, it’s pretty apt for the type of work that I do. So just to be clear, I think that networks do it, but that society makes networks. 

Let me explain with an analogy. I like to think about networks, especially networks of violence, like an interstate highway system. The system gets built over time, sometimes with good plans, sometimes with bad plans. It’s built with certain purposes in mind: which places are you going to connect, are you going to destroy a given neighborhood to open up access, etc. And once that structure is in place, those massive, six-lane highways, it is really hard to create a whole new system. Once it’s in place, that’s what you use to get around. Sometimes you can make shortcuts or new pathways or whatever, but you can’t really choose not to use the highway. When people are born, they inherit these systems. They don’t always understand the history. They just know they need to get around.

And that’s what happens with a lot of these networks where violence is concerned. They were built through patterns of housing segregation and school catchment zones and police districts and geographic political boundaries. It’s not random. There’s no randomness about why some neighborhoods don’t have grocery stores or why some neighborhoods do have lead pipes and others don’t.

When you think about neighborhoods that have high levels of gun violence and gangs or street crews, those networks help you get around. You need to know what the conflicts are, so you know how to be safe when you walk down the street, especially young people. So people navigate these networks and then they have to make decisions. If you feel unsafe, are you going to carry a gun to protect yourself? Are you going to call the police? Are you going to try to change networks? 

I do think people make choices, but the choices are severely constrained. And sometimes the choices are not choices at all. Sometimes people are saying that they feel so unsafe that they have to protect themselves, even though they know carrying a gun is going to potentially get them in trouble, or get them killed. I do think people have agency, but I think those are forced choices in some ways. 

I’ve heard you say that the average age of a gunshot victim in Chicago is about twenty-seven years old. That seemed high to me.

It surprised me the first time I saw it, but I’ve seen it consistently, which means it’s real. And it’s not just Chicago. In Evanston, the average age of gunshot victims is even older. What’s important about understanding the age distribution is that what a twenty-seven-year-old needs is not what a sixteen-year-old needs, and vice versa. I think people tend to find a young person, a teenager, more sympathetic. A twenty-seven-year-old who might have a felony conviction is more likely to be portrayed as a gang member. When we talk about today’s victims, it’s crucial to understand who they are so that we can give them the resources that they need to thrive. If you want to save the lives of gunshot victims today, you have to think about young men who are in their late twenties, who don’t have access to formal schooling systems, many of whom have their own children. 

What are the implications for policy and practice if we were to recognize this reality—that many of the victims of gun violence are not-so-young men with criminal records?

I worry about pitting short-term and long-term solutions against each other. I think when you’re talking about on-the-ground violence prevention, that network thinking can help stop cascades of violence. I think you can use this information to reach people to intervene, to prevent violence, and to save lives. I think that’s really important. But, going back to my highway analogy, if you don’t fix the structural elements, it still means the next time an outbreak happens, it’s going to be in the same place and affecting the same people. 

If the question is how do we stop gun violence today, the most important thing we need to do is build an infrastructure around the people who are doing neighborhood-level violence prevention.

Especially in the current political moment, we’re often pitting the need to address structural problems against the need to intervene in the here-and-now. The truth is that we have to do both. I don’t think we should ignore these large issues and how these systems were built. But to take those apart, whether it’s to dismantle them or to build new systems, that work is going to take generations. We have to do this work, but at the same time we have to save lives today. 

I share your belief that we’ve been confronted with what feels like a false choice between engaging in interventions to stop the violence now versus longer-term investments that might alter the structures that you describe. I’m wondering whether there are one or two examples of investments, in either of these two categories, that you think we should be making?

New York City is actually an example of a place that I think has done some things right. If the question is how do we stop gun violence today, the most important thing we need to do is build an infrastructure around the people who are doing neighborhood-level violence prevention. This means investing in the human capital and social capital of the people who are doing things like street outreach or violence interruption. How can we bolster them? What sorts of training do they need? We do a decent job of this when it comes to the police and EMT and firefighters. But our ability to support people doing neighborhood violence prevention tends to be limited and often supported mainly through philanthropic grants.

One of the things I’d like to see in almost every city is the development of a dedicated office for violence prevention—with somebody with real power overseeing a real budget—that can coordinate public safety efforts. These offices have to be properly staffed and resourced. You can’t just build these things and set them up to fail. I think New York City has done a very good job on this with their Office of Violence Prevention. Los Angeles has too. 

You can’t just invest in street outreach but then not think about schools or housing. All of those things are intertwined. But you do have to start someplace. I think having a public entity with resources coordinating violence prevention is a massively important first step.

You didn’t mention policing. If the goal is to combat a serious spike in violence right now, is there no role for hot spot policing or focused deterrence to play?

I think the research is pretty solid that policing can have an impact when it focuses on a small number of places and people and behaviors. And there are discrete models, like focused deterrence, that can be impactful when they are focused and not overreaching. 

I think police have a nonzero role in this debate. When we’re talking about gun violence, we know that they can have an impact. The other role for police, and this is crucial, is investigating. Most people can agree that we want police to investigate shootings and homicides and to solve cases. I think the problem is, as we see a surge in gun violence, people’s gut reaction is to think we need more and more police. That’s not what we want to do here. 

I want to turn to another cheeky title of yours that I liked, which was a piece you co-wrote called Why Do Criminals Obey the Law? What did you learn from asking the question in that way?

We asked the question that way in part because there’s this idea that “offenders” are somehow different, right? We were combatting the old trope that criminals believe different things than noncriminals. We already knew that wasn’t true, but what we wanted to look at was what happens if you ask them the same questions we ask the general population around things like trust in the police or belief in the law.

And so we sampled individuals who were arrested and convicted of a serious violent crime involving a firearm and we asked them, “What do you think of the law? What do you think of the police?” And what we found was that most of the individuals in our sample absolutely believed in the substance of the law. They know what’s right, they know what’s wrong. And they’re in compliance with the law the vast majority of the time. Most people that get arrested are not spending their days figuring out ways to break the law. 

Let’s be clear: their opinions of the police and the criminal justice system are overwhelmingly negative, in part because of their treatment by the system, but there’s variation. The people we surveyed could distinguish between the institution of policing versus what they had experienced personally. 

Another paper of yours was More Coffee, Less Crime?, which looked at the effects of gentrification on crime in both Black and White neighborhoods in Chicago. What did you find? 

We wrote that paper in the early 2000s, looking at how gentrification played out across neighborhoods. We used coffee shops as an indicator. The pattern was consistent: those sorts of resources emerge in White neighborhoods and not in Black neighborhoods. There’s a corollary decrease in crime in White neighborhoods that gentrify.

I think one of the things that's always hard with these kinds of programs is that you start with a small experiment to see if it works, but then when it is applied to the entire force, it's not clear that it has the same "oomph" like it once did. Scaling up is always a big problem.

Patterns of development that are often called “gentrification” are vastly different in Black and White neighborhoods. What research has shown since then is that it’s the Black middle class that gentrifies Black neighborhoods, not the sort of White hipster gentrifier stereotype. That’s another signal of the importance of race.

A lot of your work is focused on Chicago. I’m interested to hear how you think about translating ideas from one place to another. How valuable is it to compare Chicago to New York when it comes to things like street violence?

I should say, in addition to Chicago, we’ve done research in Newark, Boston, New York, Oakland, Stockton, New Orleans, Cincinnati, New Haven, and Hartford. We’ve looked at a dozen or so cities. The same three lessons—that gun violence is concentrated, that exposure matters, and that it is contagious —seem to be reproduced everywhere we look. However, network structure varies from city to city. Some cities, for example, have high-rise housing projects and some have lots of vacant, empty land. So the networks will look a little bit different, but people’s behavior within them often looks very similar.

You’ve recently turned your attention to police misconduct. I’m wondering what you’ve learned about police violence by looking at it through the lens of network science?

So it turns out police violence is a group behavior. 

Every cop I’ve ever talked to tells me the same story about their first day on the job. They come from the academy, they’re all excited, and they are paired with some field training officer or veteran who tells them, “Hey, Rookie. I know you learned all this stuff in the academy, but let me show you what real policing is like.” And then they proceed to show them all the unwritten rules of policing, including how to get away with things. At a basic level, you learn from your peers as far as policing is concerned.

Our research has shown that a small number of cops are responsible for a large number of complaints. As with gun violence in the community, exposure matters: if you’re around other cops that are doing bad things, you’re more likely to do bad things. We are learning that whether you are part of the police department or a member of a street gang, deviance is a group phenomenon. 

We’re really trying to unpack what that means because, theoretically, you have more control over policing than you do over an amorphous friendship group in a neighborhood. The police department is a hierarchy where you can administer policy and potentially change behavior. We could today, if we wanted, say, “These are the officers that are at heightened risk of shooting a civilian.” But what would you do with that information? You can’t fire them just because they’re at risk. Having information and figuring out what to do about it are often very different things.

You did a piece of work looking at the impact of procedural justice training on police use of force. What did you find?

I was not the lead author on that one, so I’m just going to speak at the broadest level. But what we found was that the procedural justice training as it was first implemented in Chicago was associated with reductions in levels of complaints and use-of-force complaints against those officers that were part of the program. It was not a massive impact, but it was not zero, either. Which does suggest that these trainings can potentially have a small-to-modest impact on outcomes like use of force.

I think one of the things that’s always hard with these kinds of programs is that you start with a small experiment to see if it works, but then when it is applied to the entire force, it’s not clear that it has the same “oomph” like it once did. Scaling up is always a big problem. 

I recently had a conversation with David Weisburd, who talked about the importance of criminologists “making the scene,” by which he meant getting out of the ivory tower and attempting to have some impact on the world of policy. You certainly have embodied this idea in your work in Chicago. I’m curious about what lessons you’ve learned from that experience.

I could not agree with Weisburd more. I think that an engaged approach to research is crucial. We have to, in my opinion, shake up how we rank or value data. The gold standard is not a randomized control trial [RCT]. You can have an RCT and not make a causal claim. And you can make causal claims without having an RCT.

Let me give you an example of something we’ve learned, which you don’t get from just looking at administrative data. We are currently working with about a dozen or so street outreach organizations in Chicago. Frontline workers are the ones doing the work, trying to mediate conflicts and engage people that are disengaged and disenfranchised. 

 How do I explain statistical power to a city council member? Do they even care? The answer is that they don't care. What they care about is, “Does it work and can you prove it?” Do I need to be fighting with city council members about propensity score matching versus synthetic control groups? No, that's stupid. Don't do that. Let's have the nerd fight at the academic conferences, let's do it in journal spaces.

What we’re working on is the idea that it’s not so much people that are risky, but situations that are risky. So risk is dynamic. Risk can change on a day-to-day, hour-by-hour basis. If all you’re doing is looking at static data to assess risk, you’re going to miss something. 

So we’re working with the frontline workers to learn from them about what kind of information they think is important. It’s not something you could get by saying, “Well, can I add one more variable to the social learning theory in my statistical model?” I mean, I suppose that’s interesting, but it’s way less interesting than really trying to understand how to keep people alive.

I’ve found that every single time I’ve engaged with practitioners in this way that there’s always a gazillion interesting theoretical things and theory-relevant things that can come out of it. But the more interesting questions are coming from the outreach workers in this case. And the only way you get at it is by engaging with them.

So we’ve codesigned interviews and we’ve built an entire survey instrument with our outreach partners. We sit down and analyze data side by side. They are able to provide insight into what our findings mean. And when they get interested in something, we go deeper. 

I think it’s important to recognize the power dynamic though. I mean, I’m a researcher at an elite institution and I’m working with nonprofit organizations that are struggling to keep the lights on. And so it’s important to also understand the footprint of the criminologist in the field, especially as you’re trying to answer questions that may impact funding.

Have you paid any professional price for your engagement in the “real world”?

I’m fortunate enough where I’m at a stage in my career that it doesn’t impact me in the same way [it would] if I were a more junior scholar. I’m able to take risks.

I do get dragged into a lot of academic debates around the value of “observational data” and whether it is somehow lesser. I think there is an idea in our field that somehow observational data are bad. I think that’s harmful to science actually. It’s also harmful for the communities that are affected by gun violence. 

Gun violence is not random, so why do we pretend like it is? Once you start to see these networks, you can’t unsee them. I can pretend like they don’t exist in a regression analysis, but I know they’re there. The people who are involved with gun violence, they know each other. They went to school together, they’ve got family relationships. So why are we pretending like they’re not? Can’t we amplify and boost that understanding? We think somehow our findings are lesser because there’s not a statistically significant star at the end of the equation. This bias towards certain types of causal logic stifles innovation. 

And as far as public policy is concerned, the bar is set in such a way that most of the programs we evaluate will always fail. In Chicago, some of the programs we’re evaluating now are reaching a population that’s hard to serve, so they’re working with a few hundred people. Well, you are never going to get a statistically significant finding with those kinds of numbers. It’s not going to work because you don’t have fifteen hundred people in your sample. But how are you going to find a program that can service fifteen hundred people with the types of budgets they have? You’re not going to. I think those tensions really are stifling creativity and knowledge in this space. I think if more researchers got out there we could probably advance the field.

Two of the things that I’ve heard from talking to other researchers about their engagement with the world outside of academia are a fear that their work might be misused and a concern that, when dealing with the media or with politicians, that they will have to sacrifice the nuance of their work.  Have you had to confront either of those things?

When it comes to the idea of your research being misused, I’m somebody who that’s happened to on multiple occasions. If you believe in open science, if you believe in sharing your ideas and your data, you’re always going to be open to being misused. The path I’ve taken is to push back when this happens. I’ve written op-eds about the Chicago Police Department taking ideas of mine and incorporating them into projects in ways that I thought were horrific. I have also critiqued the Cure Violence model for doing the same thing. I’m not going to just shrug my shoulders and say, “Oh, there’s nothing you can do about it.” But at the same time, I think it’s really vital that science get out there. 

I think one has to be clear on what you think should be done, which is what I’ve tried to do. I think we should use network science stuff for on-the-ground violence prevention efforts, not for arrest-driven police behavior. That’s an important distinction I’m consistent on. 

The nuance question is really tough. I do not think we should abandon nuance. I think we need to train criminologists how to write better. I do believe it’s crucial to produce a document that has all the nuance in it. But when you get in front of City Hall, when you get called to testify before such-and-such committee, when you’re talking with a local nonprofit that wants to understand how this research will help them, you have to be able to say it in a couple of bullet points, and those bullet points have to be translatable to action. 

I’ll give an example. When we’re talking about street outreach efforts in Chicago, there are some very clear findings: One, you can find the right people [who are engaged in violence]. Two, you can connect those people with services. And three, those individuals basically do better in terms of outcomes like reduced victimization and violent arrests. Everything I just said is true. Here’s the nuance: it is not always statistically significant. Sometimes it is. Sometimes it’s not. How do I explain statistical power to a city council member? Do they even care? The answer is that they don’t care. What they care about is, “Does it work and can you prove it?” Do I need to be fighting with city council members about propensity score matching versus synthetic control groups? No, that’s stupid. Don’t do that. Let’s have the nerd fight at the academic conferences, let’s do it in journal spaces. 

Some academics don’t seem willing to even entertain the idea that police could ever reduce crime. To me, it feels like they are starting from an ideological place and not looking at the evidence. Do you think that this is happening or am I misrepresenting what’s going on in your field?

The first thing I tell graduate students is, “You can’t ask questions you don’t want answers to.” The questions you ask are going to put you on one side of something or other, and you better be prepared for the answers. Most often than not, to go back to your nuance question, the answers are super complicated. 

We are just wrapping up a project on neighborhood policing in Chicago, where we interviewed police officers, community residents, and community residents that were less engaged, ones who didn’t show up to the meetings or weren’t part of any particular organization. So we interviewed these groups of individuals every three to six months for two years before George Floyd was murdered and we’re still interviewing them now. Even before 2020, what we were seeing was the variation and complexity of people’s opinions about public safety and policing. And it gets even more complicated in 2020. What we found is that both police and residents can differentiate between individual people and institutions. So I can like officer Greg, and still say the Chicago Police Department is a racist institution. I can say that I want to change the CPD, but don’t take officer Greg away because he’s the only one who gets me. In people’s minds, that’s not a conflict. They can hold those two thoughts in their mind at the same time. So, the on-the-ground view is really complex. 

I think there’s variation among academics. I think some academics are picking and choosing the questions they are asking based on where we are in terms of the current political moment. I don’t think that’s entirely bad. I do think that this is a long game, right? Crime and violence, policing and public safety, these are not new problems. So I think it’s good to take up new perspectives and ask new questions from an academic perspective, but you have to be willing to understand the answers, even if it doesn’t go the way you hoped it would go.

As an example, I would love to get up and say that street outreach is the most impactful thing we can do to reduce gun violence today, but I can’t say that. I can say it’s super promising. I can say that sometimes we see evidence that it works, but I can’t say that this is the solution to gun violence. I can’t say that, even though I personally really want to. But as a scientist, I can’t say that. As a scientist, I have to say, “Here’s what we know and here’s what we don’t know.”

Greg Berman is the Distinguished Fellow of Practice at The Harry Frank Guggenheim Foundation. He previously served as the executive director of the Center for Court Innovation for 18 years. His most recent book is Start Here: A Road Map to Reducing Mass Incarceration (The New Press).

Views expressed are the participants’ own and not necessarily those of The Harry Frank Guggenheim Foundation.

Previous At the Crossroads interviews:

Welcome to the website of The Harry Frank Guggenheim Foundation

Sign up here for Foundation news and updates on our programs and research.