310. Community Effects of Incarceration

Some students receive substantial support on their educational journey within their homes, communities, and schools; others face substantial barriers. In this episode, Arpit Gupta joins us to discuss his recent study that examines the effect of community incarceration rates on the academic performance of children in affected households and on their classmates.

Arpit is an Associate Professor of Finance at the Leonard N. Stern School of Business at NYU. Arpit has published extensively in highly ranked finance, economics, science, law, and management journals on topics ranging from housing markets, infrastructure investment, bail, local journalism, racial housing gaps, incarceration, and remote work.

Show Notes

  • Gupta, Arpit and Hansman, Christopher and Riehl, Evan (2022). Community Impacts of Mass Incarceration. May 3.
  • Norris, S., Pecenco, M., & Weaver, J. (2021). The effects of parental and sibling incarceration: Evidence from ohio. American Economic Review, 111(9), 2926-2963.
  • Lazear, E. P. (2001). Educational production. The Quarterly Journal of Economics 116(3), 777–803.
  • Chetty, R. (2016). Improving opportunities for economic mobility: New evidence and policy lessons. Economic Mobility Research and Ideas on Strengthening Families Communities the Economy, edited by Brown, Alexandra, Buchholz, David, Davis, Daniel, and Gonzalez, Arturo, 35-42.
  • Chetty, R. (2021). Improving equality of opportunity: New insights from big data. Contemporary Economic Policy, 39(1), 7-41.


John: Some students receive substantial support on their educational journey within their homes, communities, and schools; others face substantial barriers. In this episode, we discuss a recent study that examines the effect of community incarceration rates on the academic performance of children in affected households and on their classmates.


John: Thanks for joining us for Tea for Teaching, an informal discussion of innovative and effective practices in teaching and learning.

Rebecca: This podcast series is hosted by John Kane, an economist;hellip;

John: ;hellip;and Rebecca Mushtare, a graphic designer;hellip;

Rebecca: ;hellip;and features guests doing important research and advocacy work to make higher education more inclusive and supportive of all learners.


Rebecca: Our guest today is Arpit Gupta. Arpit is an Associate Professor of Finance at the Leonard N. Stern School of Business at NYU. Arpit has published extensively in highly ranked finance, economics, science, law, and management journals on topics ranging from housing markets, infrastructure investment, bail, local journalism, racial housing gaps, incarceration, and remote work. Welcome, Arpit.

Arpit: Thanks so much for having me.

John: It’s great to see you again. It’s been a while since we last talked… 20 years or so.

Arpit: Yeah, it’s been a while. So I owe my economics career to John having him teach me at a very formative time in my life. Very happy to be back here.

John: Back at the TIP program, way back. And you would have probably done that anyway, because you had a lot of interest in it even back then. Today’s teas are: …are you drinking any tea, Arpit?

Arpit: …just drinking water at the moment.

Rebecca: It is the foundation of tea.

John: It’s one of our more popular teas.

Rebecca: I have an Awake tea today.

John: I have a Darjeeling tea today.

Rebecca: So we’ve invited you here today to discuss your May 2022 working paper on community impacts of mass incarceration, co-authored with Christopher Hansman and Evan Riehl. Could you tell us about the origin of this study?

Arpit: Yeah, so Chris, Evan, and I were all graduate students at Columbia University. Chris and I were also roommates. And we had a third roommate who was a public defender. So we would just come home and hear interesting stories of his experience at work and things he was seeing. One of the things that he brought home and kind of talked to us was the fact that bail was an interesting process. And there was an interesting random assignment across bail judges. And so that was our first project, it kind of stemmed directly from talking to this roommate and his collaborator on that project. And another thing that he was mentioning is that the way he saw it is that incarceration spells really had rippling effects, not just directly on individuals concerned, but kind of affected broader communities in different ways. And we felt that that was a really interesting insight that has been explored in some other non-economics research. And we wanted to just explore this concept further, because we felt it was an important essential public policy question. And so we spent many years to try to get the right data and setting to explore further at these broader community impacts of incarceration.

John: So earlier studies had found that incarceration of a parent had significant effects on education for children within the household. Could you just talk a little bit about those effects before we talk about your contribution to this literature?

Arpit: Yeah, absolutely. So there is a pretty broad literature on this topic. And I would sort of separate some of the papers that are not in economics from the papers that are in economics. There are a number of great studies that, for example, will track cohorts of people across generations to kind of see what are the rippling long term implications of incarceration. There are a variety of these papers that explore I would sort of describe are the multi dimensional aspects of incarceration on different outcomes for individuals and families that are concerned. And I would sort of characterize this non-economic literature as really highlighting the disproportionate spatially concentrated incarceration. And that’s kind of the key insights of this broader sociological literature, that you think of incarceration as something that affects a lot of people in very concentrated ways and bad ways. The economics literature has taken a little bit of a different approach and has primarily focused on the direct impacts of incarceration, with some literature starting to look at how that also affects household members. A lot of literature has been in Scandinavian countries where they have a different justice system and really good data. Some of those papers have actually found positive effects of parental incarceration on children outcomes, which might make sense if you’re removing, for example, a negative role model from a child’s life or if the criminal justice system itself offers positive remediation, restorative justice, and so forth, that kind of improves someone’s outcomes after they’ve returned from prison. The closest paper to our study in the United States is going to be a paper by Norris and Weaver, which focuses on the effects of incarceration for students in Ohio. And there, they argue that incarceration of a parent improves the odds that the child is going to be involved in the criminal justice system in the future, so that they are less likely to be arrested in the future. And they find more mixed evidence on the education impact. They don’t find much evidence for negative education impacts. But that’s done on the kind of little bits of a smaller sample with larger standard errors.

John: Your study, though, goes a little bit further, because you’re looking not just as the effect on children within the household, but also spillover effects into their classrooms and schools, from incarceration of adults in the household. How did you separate out the effect of differences in incarceration rates from all the other factors that might influence such outcomes in those communities?

Arpit: Absolutely. So this is going to be, of course, a key distinction between how economists think about the problem versus other disciplines. We’re thinking upon the question of identification. How can we identify whether the negative impacts or positive impact you’re looking at can be attributed to incarceration, or are just reflective of other background trends. Let me start first with actually how we think about these effects in aggregate, because that gets at like the community dimension of the problem, which is kind of our central focus. So the big question that we’re really interested in is what happens to a community, when a lot of people within that community are behind bars? How does the impact on that set of individuals spill over and impact the overall community. And of course, this is an even harder identification problem than just looking at the attacks on one person, because you wonder what the omitted background factors that can affect entire communities. But we find that when a county has a relatively more strict set of judges, that actually has a large impact on the overall performance of all the students in the area. So that suggests that there are large impacts of incarceration that kind of broadly affect all the students in a particular area. And that motivates us to think about what is the size of the effect of incarceration on children’s outcomes, and what are the mechanisms by which they’re affected? But we then dig more deeply into thinking about the effects on the directly affected children, those whose parents are themselves incarcerated. There, we similarly use judicial variation, and we also look at the spillovers onto other children in the classroom. So the key innovation, the key contribution, I think, of our analysis is to take this question that has been studied before, but adopt it to the problem, thereby thinking about the more aggregate consequences and the mechanisms by which incarcerations affect broader communities.

John: And you also use an event study approach too, to provide more support. Could you talk a little bit about that part of the analysis?

Arpit: So we use those in both our direct and indirect analyses where we were trying to understand what is the impact on a student if their family member is incarcerated. And the event study approach basically looks before and after that arrest and looks at the outcomes for the children as measured by outcomes such as the test scores, the suspension rates, misbehavior rates, and so forth. So we’re interested in a little bit of a multi dimensional set of outcomes for children, because we want to know both how is this child doing, we want to know whether there are behavioral disruptions that may stem from having a background incarceration at home, that may then affect other children, because if you’re misbehaving in the classroom, that’s something that will negatively potentially affect other children’s learning in the classroom. The event study is looking within the child before and after that arrest period. And we also do that same event study analysis at the classroom level, basically. So looking at what happens to the performance of other students in the classroom, when one of the students’ family members is arrested.

Rebecca: How big was the impact of incarceration on children in the affected households and in the classrooms.

Arpit: So for one individual child, the effects on math and English scores is something 5% of the standard deviation. So it’s an effect that is sizable enough, if you think about many educational interventions as having very heterogeneous effects, and it’s very hard often to kind of get meaningful moves in child performance. But the really big part of the analysis, I think, was trying to reconcile those direct effects, the ones that are one to one and a half percent of the standard deviation against the overall impact of incarceration on the whole community. So what happens if I take a whole county and I change the mix of judges and I have much more incarceration? What is the overall educational impact there? So when we looked at that overall community level perspective, we actually found that changing a one standard deviation in the county level stringency is actually affecting test scores by between one and a half to three and a half of a standard deviation. So we’re basically getting very big aggregate effects that the individual effects alone can’t explain. And so we think that there’s scope for these spillover effects, by which one directly affects how a child behaves in a certain way in the classroom that then spills over to the other children in the classroom that thereby amplifies the effect, so as to generate larger negative overall effects. And one channel that we use to identify those is to look within the classroom itself, not gonna measure all the potential spillovers between children, but it’s one area where we think there’s spillovers, and we think that those spillovers can also account for some fraction of the overall community effect.

Rebecca: Can you translate some of that standard deviation talk [LAUGHTER] to people that don’t know anything about statistics.

Arpit: For example, at the county level, when we are thinking about a one standard deviation increase in the stringency we’re thinking about a 15 to 20% increase in incarceration. So that’s kind of the range of variation that we’re looking at at the county level when thinking about what are the typical shock to incarceration, and that’s a kind of pretty substantial increase in the incarceration levels we’re seeing as a consequence.

John: So you’re finding the effect on any one other student is relatively small, but the aggregate effect on all the students in the class is relatively large. Is that correct?

Arpit: That’s right. So when we look at those other students in the classroom, we’re getting effects for those students in response to the incarceration of a peer’s family members, they’re on the order of 0.3, 0.4 percent of a standard deviation. You should just basically think of that as a really small number. And the only way we’re kind of getting the power to analyze this is that we’re looking at this North Carolina data, which is really great, a lot of people have worked with it, exactly because it is so comprehensive. So we’ve got all the student rolls, we’ve got all the arrest records, all of these are matched together. And so using this really holistic sample allows us to try to quantify these effects that are pretty small for any one individual child, but they’re just a lot of exposures that can aggregate up. And so we think that this classroom disruption channel can explain something like 15% of that relationship between aggregate incarceration and test scores. So it kind of all adds up to explain a more meaningful fraction of this overall relationship between what happens when a lot of people in the area go to jail and what happens to student performance in that area.

John: What sort of mechanism are you hypothesizing might be the cause of the spillover effects to other students in the classrooms?

Arpit: So let’s start with what we can measure in our data. So what we observe is that children who are affected whose family members are incarcerated are looking at increases in suspension days, they’re absent more often, they’re involved in more fighting incidents, typically it takes two people to fight. So that sort of tells us that there are other people involved in the classroom for these affected students. And so we think that this relates very closely to the idea that there are classroom level externalities, and there is a large literature, actually papers by Lazear and others that highlight the importance and implications of classroom level externalities, classroom disruptions, when it comes to learning. It also comes up, by the way, when I talk to people in North Carolina who are teachers. One thing that they really bring up is that children come into the classroom with all sorts of backgrounds that change behavior in the classroom, and that impairs the learning experience for other students in the classroom. So that’s what we can measure most cleanly, is the existence of these behavioral disruptions by students affecting how they behave in the classroom, and influences, through that channel, the learning experience of other children. That doesn’t need to be the only mechanism that’s going on here, there can be other spillover channels between children that we can’t observe in our data. There can also be other channels outside of peer interactions between children through other community interactions between people as well, that we also can’t measure in our data. So we think of this project as really trying to open a set of analysis that we’re considering and thinking about the broader web of social interactions, when incarceration happens.

Rebecca: What are some of the public policy implications of the study?

Arpit: So the challenge, of course, is that you’re measuring one side of the equation, we’re measuring sort of the cost of incarceration, and so you have to balance those against some of the possible benefits of incarceration, because children are also affected by crime in the local community, as well. And so it’s a difficult trade off to try to balance both the costs and benefits of incarceration in tandem. So I don’t think our results actually have a clear takeaway. I think the biggest thing that I personally kind of took away from the analysis is that if we have different techniques, if we have different ways of trying to reduce and address crime, it would be ideal if we were able to lean on ways that rely less on the incarceration channel, which impose these additional externalities and costs and burdens on local communities, and instead found other ways of trying to address and mitigate and reduce crime. So for example, when it comes to a different setting, when it comes to thinking about bail, which is a topic we’ve also researched before, there is sometimes a choice between arresting the individual and putting them in jail, compared to something like house arrest, compared to something like electronic or digital monitoring. These systems are also not perfect. There are also a lot of costs and tradeoffs there. But to the extent that you can find ways of deterring, mitigating, crime that don’t rely as much on the incarceration channel, I think that lowers the spillover negative effects on local communities, I want to mention that, when we look at these multi-dimensional impacts of the original incarceration event on the student, we actually find, consistent with prior literature, that to the extent that we can observe juvenile offenses, we don’t observe increases in crime, if anything, there are decreases in criminal activity. That, again, is consistent with some of the prior literature. And the way to interpret that, I think, is to again think of there as being multiple dimensions by which people are affected. So you can observe that there’s a negative role model effect, you observe someone going to jail for a crime… Well, I’m not going to commit that crime, but you may still act out in the classroom. So we shouldn’t think of the responses to these kinds of disruptive background events as happening on some uniform spectrum of good behavior or bad behavior, but it’s much more multi-dimensional in how people respond to stressful situations.

John: Did you find a difference in the effect whether it was a male or a female in the household who is incarcerated in terms of the impact on children?

Arpit: Everything I’ve said, so far, I’ve been trying to be careful in sort of saying , these are individuals in the household, because really, what we’re doing is the household level match. So we’ve got the address, and so what we really know is that this is someone that lives at this address that is arrested. We view that mostly as a strength of our approach, which is trying to identify household members. It sort of recognizes the intergenerational and complicated family backgrounds many families have, but it does make it a little more challenging to establish the sort of true relationship between individuals. And so one thing that we kind of did there is sort of try to identify probable female parents or guardians, male parents or guardians, or simply assign kind of age ranges and things like that. We did find the effects on children were much larger when we were looking at the incarceration of a female payment. So that kind of makes a lot of sense, if you think that mothers and female guardians kind of play uniquely important roles within the household. And when it comes to the child themselves, the effects were actually pretty similar between boys and girls.

John: In the US, we have one of the worst rates of intergenerational income mobility, might this type of an issue be one of the causes of that, in that in low-income communities where incarceration rates tend to be higher, it’s putting children in those communities at further disadvantage, which can have some long-term consequences.

Arpit: One thing I want to mention is where we’re kind of taking the paper is to adopt the community frame and think about other community outcomes that might potentially change as a result of incarceration. So I do think that probably one of the reasons that we have this, not just low on average in the United States, a low rate of social mobility in the United States, but also it’s very regionally varying rate of social mobility differences across the United States. I remember when the first Chetty map was released that showed the geography of economic mobility in the United States. My home state, North Carolina, is actually incredibly low for social mobility. And that’s surprising, actually, because North Carolina is where everyone’s moving to. It is incredibly economically dynamic, it has lots of job centers, but moving there is low cost of housing. It has a lot of features, which you might expect should be associated with high economic mobility. And in fact, like much of the south and very regionally varying patterns across United States, you actually observe pretty low social mobility. And I do wonder whether one reason for that is that we have these very high rates of incarceration across much of the United States. And that’s not an easy thing to just stop incarceration, because we all know that the system of criminal justice, that is also there to protect in low- income communities from the negative consequences of crime. So the public policy challenges of how to figure out what to do about this are really complicated. But we want to know why is it that people that grew up in the same state that I did, don’t necessarily have great opportunities compared to people who grew up elsewhere. So we’re hoping to use the setting, use this analysis to dig a little bit deeper into this question. And one fact that is kind of already out there that I think is very related, is that analysis by Chetty and others, which looked at the geography of social mobility, found that a big correlate, something that associates strongly with social mobility across United States is the presence of two-parent households. So the number of absent fathers, that associates very strongly with the lack of social mobility in an area. Of course, that is not a causal statement, you could imagine things go the other way. So lack of social mobility kind of impacts in different ways. But I think that’s a diagnostic that is suggestive of the idea that something about incarceration affects broader communities, affects the family formation, affects family stability in ways that impact people’s ability to build stable relationships. And all of that kind of has really persistent negative impacts.

Rebecca: As an educator, this study makes me think about if I’m a teacher in a classroom, I’m kind of experiencing the phenomenon that you’re studying, and the kinds of things that I might consider doing for classroom management or the way that I might better understand even just what’s happening or what I’m observing, I think is food for thought for educators to just be more aware of what’s happening in their communities.

Arpit: The other kind of question, I think, for economic policy is about these measures of teacher value add, which are being thought of as ways of assessing or even compensating teachers for the increase in test scores, that they’re resulting in the classroom that they have, right? And this makes sense to economists we want to value and grade people based on the incremental add that they’ve done to a population kind of coming in. But one thing I actually hear a lot from teachers is they’re very worried about this possibility as something that affects them as a teacher, because they’re saying, “Well, it’s not my problem, if I happen to have a classroom in a particular year where the children are going through a lot of stuff at home, they’re not necessarily going to learn as much, that might affect other children in the classroom as well. And that’s something that I will potentially be judged for, something outside of my control.” And that is a very strong problem for this whole teacher value add methodology, because these kinds of background events don’t necessarily follow a predictable sequence. And so they can kind of happen at various times over students’ lives, over a teacher’s career across different classrooms. And so it’s very hard statistically, to separate out whether a student is doing well or badly because of the teacher, or because of some background events. It also impacts, I think, how we statistically evaluate and think about evaluating teachers.

Rebecca: I imagine it also impacts classroom management and observations of classroom management and other tools that we use to evaluate teachers currently… behavior in that class is different than others, or they have different traumatic experiences impacting their behavior. That’s not necessarily being observed by an observer.

John: And we have put probably far too much weight on teacher and school compensations and budget tied to student performance, because, as you said, there’s so much that’s outside the control of the teachers or the school districts.

Rebecca: That’s also the schools that tend to struggle to get teachers and things too, right?

John: And we’re penalizing those teachers and those school districts, often, that face the most severe challenges and need the most support. You mentioned this dataset from North Carolina is a very rich one, but you had to do a bit of work to get all that data together. Because there is a lot of data on student outcomes, but you also have to tie this to incarceration. Could you talk a little bit about how you matched the household data, or the incarceration data, to the schooling data?

Arpit: Oh, man, this is my favorite part of the project, because it allows me to reminisce about my sort of a Moby whale moment of a project. So I think all of us as researchers need to sort of think about what are the projects that we really want to see live, what are the ones that we’re really going to go to bat for, and this is one of those projects for me. I just felt that this needed to be answered. And so, together my collaborators, we really just spent a really long time trying to figure out how to get the right data for this. So you have to put together the criminal justice records for a given area, you need to put together the education records, and then you need to figure out how to link the two of these. So some states you can get one, some states you can get the other and it’s very hard to find a set of states where the two of them match. So we tried a whole range of states, a whole range of datasets, many times we got very close, but were stopped at the last minute. And finally, we were able to work with the state of North Carolina, which has an excellent set of education records, has these great criminal justice records, and were able to figure out a way of merging and matching the two sets of documents at this household level, have a pretty good sense of the direct linkages between the children in our sample and the criminal defendants and then using the classroom identifiers in the dataset to identify other spillover effects, looking at the broader geographic implications. So all of that wound up working out for us at the end, but it was a long haul to get there. And I think it’s definitely a lesson that I took away from this project that if you want something to do well, you really got to work at it. There’s no substitute for putting in the shoe leather for calling people, cold calling people, emailing people and just hearing no, no, no, no, no again and again and again until you’re able to figure out something that works.

John: And the matching between households for the students and the incarcerated people was based on household addresses. Is that correct?

Arpit: That’s right. So that match was done by the North Carolina education folks. They took their records, they imported the criminal justice records, matched that at the household level, and then gave us a data set that had removed all identifiers that we could work with for research.

John: It’s a wonderful data set and it’s a really impressive piece of work.

Arpit: Thank you very much.

Rebecca: So we always wrap up by asking what’s next?

Arpit: For us on this project, we’re really trying to see if we can think about some of these broader implications of incarceration on communities outside of the educational impacts that we’ve been talking about so far. So thinking about the impacts on family structure, thinking about whether it spills over into the usage of other government programs, whether it has employment effects, kind of housing market access, I think that there are a whole range of different outcomes, particularly at these broader community levels that I think are shaped by the number of people in that local community that are impacted by incarceration. So I think those are the overall community spillovers, we’re interested in understanding.

John: Well, thank you. This is some really impressive work. And I have to say I’m really impressed by all the work that you’ve been doing in so many areas. You’re doing some wonderful work on some really important topics.

Arpit: Thank you very much, John. I had an economics teacher growing up who inspired me to work on these topics.

Rebecca: Well, thank you so much. We’re looking forward to sharing this with our audience.

Arpit: Thanks.


John: If you’ve enjoyed this podcast, please subscribe and leave a review on iTunes or your favorite podcast service. To continue the conversation, join us on our Tea for Teaching Facebook page.

Rebecca: You can find show notes, transcripts and other materials on teaforteaching.com. Music by Michael Gary Brewer.

Ganesh: Editing assistance by Ganesh.