Probable Causation

View Original

Episode 7: Jeff Weaver

Jennifer Doleac

Jeff Weaver

Jeff Weaver is an Assistant Professor of Economics at the University of Southern California.

Date: July 9, 2019

A transcript of this episode is available here.

See this content in the original post

Episode Details:

In this episode, we discuss Professor Weaver's work on the spillover effects of incarceration:

"The Effect of Parental and Sibling Incarceration: Evidence from Ohio" by Samuel Norris, Matthew Pecenco, and Jeffrey Weaver.

OTHER RESEARCH WE DISCUSS IN THIS EPISODE:

Transcript of this episode:

 

Jennifer [00:00:06] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an Economics Professor and the Director of the Justice Tech Lab.

 

Jennifer [00:00:18] My guest this week is Jeff Weaver. Jeff is an Assistant Professor of Economics at the University of Southern California. Jeff, welcome to the show.

 

Jeff [00:00:26] Hi, thanks. Thanks for having me on the podcast.

 

Jennifer [00:00:28] So we're going to talk today about your recent work on the effects of incarceration on the children and siblings of those who are sent to prison. Could you start out by telling us about your research expertise and how you became interested in this topic?

 

Jeff [00:00:42] Sure. So broadly speaking, I'm an applied microeconomist with a particular interest in crime and illegal markets. So this is a pretty broad research interest, some of my past research has looked at illegal markets in the developing world such as corruption, dowry in India, ways of reining in corrupt government officials. But more relevant for this podcast, I also work on crime related topics in the US, such as a project we're talking about today on how children are affected by incarceration of family members. So here, how children are affected by incarceration of mothers, fathers, as well as their siblings. And so I started out becoming interested in this topic during graduate school. I was reading a lot of sociology, particularly ethnographic work, focused on lower income communities in the US. And this was partially a way to get a little bit closer to the real world, a little bit away from the math of grad school. But through a few books in particular, focused around the the criminal justice system, it just really jumped off the page how common it is one for people to have incarcerated family members, and then two just the different effects that this could have. And when I started looking outside of the more qualitative work and to more quantitative work for for research on these effects, it just didn't seem like we had that much in the way of really convincing causal evidence on what the effects of incarceration were, not on the person who himself is incarcerated, but on the other people in their life, such as their family members, community members, etc. And generally it seems like these these spillover effects may potentially be pretty consequential. So if a child has a caregiver removed from their from their life, it's pretty easy to imagine some negative consequences from that. So emotional trauma or potentially as economists, we might think this is removing resources from the households. This is not just economic resources, but also caregiving resources and this potentially may be quite negative.

 

Jeff [00:02:29] At the same time, though, it's not necessarily obvious this is going to be the case. Where in some cases, it seems like removing a parent who, or a sibling, who is at the margin of incarceration could actually potentially have some positive impacts. For one, it could cause a child to transition into a more stable home environment. I think the most obvious case of this is when the crime that the parent is accused of is one of, say, abuse against the child. It's pretty easy to imagine that this is a reason- this is a case in which it may be that removing this parent from the child's life temporarily may actually be beneficial if this allows them to, say, shift and and live with potentially a more stable set of caregivers such as potentially grandparents, aunts and uncles, etc. And just more generally, it seems like something we'd really want to know more about as there's a really large population that's being affected. So there's one study that found that there's over five million children in the US who have had a parent incarcerated at some point in their lifetime. And so because of that, just given that there's this huge population be affected being affected. Given that it's not necessarily clear what the effect of this is going to be and that it may be the case, this is actually pretty heterogeneous where there's some children who are going to benefit from this, and there's other children who potentially are going to be harmed by this, just trying to understand what these what these effects were gonna be just seemed like something that'd be quite useful for us to know. So that's a bit on how I came to it. Though I should also say, as you said, this is this project is joined with two really amazing coauthors, Sam Norris and Matt Pecenco, where actually all of us had started working on this when we were in grad school. We learned the others were working on it and then ended up deciding to work together. So that's- so I actually- so they came to it in slightly different ways, but that's that's really how I came to it.

 

Jennifer [00:04:09] That's a great story about overlapping research leading to a positive outcome. People tend to freak out when they hear that people are working on the same paper they're working on. It's good when you can combine efforts. One of the things that I love about working on on crime more broadly is that there's there's a lot of bipartisan interest in reducing our dependence on mass incarceration in the United States and thinking about finding more effective and efficient ways to reduce crime. And I think this is motivated largely by a sense that what we're doing currently is just not efficient. But it's also motivated in part by this concern that you mentioned, that incarceration has important negative externalities. That is that locking people up can have detrimental effects on their kids and on their communities. That those detrimental effects might cancel out some of the crime reduction benefits we receive. And so your paper is is one of the first to really try to quantify those spillover effects in a rigorous way. But I'd love for you to set the stage for us a little bit. You mentioned that there's there's a lot of sociology work on this. What had been previously known about the effects of incarceration on the families and communities of those who were incarcerated?

 

Jeff [00:05:14] Definitely. So, there's a lot of research on this, and I guess you can sort of split it up into a few broad groups of evidence. So the first, as you said, is more qualitative evidence, particularly in sociology. So this is one we often skip over as economists, but as someone who personally really loves sociology research, I think we lose out a lot by it, by not looking at that. And so what most of these studies are looking at is they're following children who have incarcerated parents and trying to understand their experiences. So there's one book that I really like by Jane Siegal, who's a criminologist at Rutgers, called Disrupted Childhoods. And so this follows around 70 children who have incarcerated mothers and tells their stories and experiences. And so I think this gives a pretty rich view of what's happening. And something I think's really nice and really interesting in this is that you see that there's really a range of different experiences where there's some children for whom this is really emotionally traumatic. And you can see the just through their stories, I mean, this isn't quantitative evidence, this is more qualitative, but it seems relatively convincing. This has some negative effects on them later in life. But she also presents other cases in which the children are really in a home situation, which is very disruptive, particularly in cases where their parents may be suffering from substance abuse issues or other related issues. And it may be the case that in these and some of the cases she looks at, this is the case, it was actually potentially beneficial for the children by moving them from unstable living situations. And so I think that's a really useful piece of evidence. The issue with that in terms of thinking more quantitatively is it's hard to know, one how representative the sample of 70 children is. Though, I think she's really to be applauded for not just looking at, you know, four or five children, but trying to look at a very large sample. But even so, even if this is a pretty representative sample, it's kind of hard to know how to aggregate these, though. So what's the net of- so if you aggregate across all 70 of these children, what's the net effect going to be? To know that you really have to know how is- how representative or unrepresentative the sample is, and so it's harder to tell a more a more broader, more quantitative picture.

 

Jeff [00:07:13] In sort of the more quantitative evidence, there's a really large body of work, again, mostly in quantitative sociology. So there's probably at least over 100 papers I would say, I'm not quite sure how many in all. But what these papers are mostly doing is using survey data. And so they'll have survey data where there's some children who have incarcerated parents, other children who do not. And then they'll compare between the children with incarcerated parents and those who do not and look at later life outcomes. So these could be things such as performance in school. These could be things like labor force participation. Even there's a number of papers looking at health effects of having incarcerated parents. But sort of the issue with this, is that we might think that children who have incarcerated parents are probably going to be quite different from children whose parents are not incarcerated. One, in ways that are potentially possible for us to measure. So maybe their parents will have lower levels of education or maybe be less wealthy in ways that we can measure. There's also a lot of other ways in which it's going to be- in which they're going to differ which is gonna be pretty hard for us to to measure and to try to control for. So wealth, for example, is gonna be pretty hard to control for. Neighborhood characteristics may be hard to control for. Potentially the psychological aspects of being in a situation where parents are on the margin of incarceration are potentially gonna be hard to control for. And so these papers, which are making these comparisons and trying to control for these other differences, try to only cover the effect of parental incarceration, potentially, there's gonna be a problem with omitted variables where there's a lot of unobserved things that's gonna be hard for us to fully control away. And so most of these papers find relatively negative effects of parental incarceration, but it's a little bit hard for us to know is this because of parental incarceration having a causal effect- a causal negative effect on the child's later life outcomes or is this just because we haven't been able to control for all the many types of disadvantages that children with incarcerated parents tend to have versus children whose parents are not incarcerated?

 

Jeff [00:09:13] There's also some more more recent work done predominantly by economists that tries to look at this question more causally. So one style of paper that's looked at this is using more of a difference in differences style approach. So there's a really nice recent paper by Steve Billings using data from North Carolina schools where what he does is he looks at the set of children whose parents are arrested or incarcerated and in effect is comparing those children's test score outcomes and behavioral outcomes, comparing the times when their parents are incarcerated or just been arrested versus the times in which they were not. And he actually finds relatively positive effects of having arrested parents on on children's test score outcomes and behavioral outcomes. And so that's sort of a third body of work. And there's a few other papers that are doing something similar.

 

Jeff [00:10:03] And then finally, there's a bunch of papers that are doing something very similar to what we're gonna be doing and we're gonna be looking at more of the long run effects of parental incarceration. So papers like the Billings paper is really nice for looking at short run effects. So you can see right after the parents are arrested, do we see that children's test scores tend to go up? Do they tend to go down? Do they tend to stay about the same? But something else we might want to know is what's going to happen 10 years down the line or 20 years down the line or 30 years down the line. And so there's five papers that have come out in the last year or so, including ours, that are using using a slightly different strategy to try to get at what is this long run effect? But trying to get at the causal effect of incarceration and try to get around this omitted variable bias problem that we have when we're just comparing children whose parents are incarcerated to those who are not and then trying to control for differences between them.

 

Jennifer [00:10:54] So as you said, there's been a bunch of new papers that just came out in the last year or so, and that's been really fun to see. But it's always interesting, like you know, why now? It is kind of the question that comes up. And so as you and your coauthors were approaching the study and brainstorming ways to get at this question, what were the main constraints that you had to overcome? Were they identification challenges? Were they data challenges? Was it both? What were the hurdles here?

 

Jeff [00:11:20] Great question. It's definitely both of those. And I think that also speaks to why all of these papers are coming out at around the same time. So I guess I started working on this project July of 2015, I think it was, so about four years ago. And I think all the other all the other folks working on these other projects have also been working on these on these papers for a huge amount of time. And the reason for that is just it takes a really long time to try to put this data together where I think a lot of people had been thinking about this and thinking, OK, so there's potentially some strategies that we could use to get around this. But it takes a really long time to put this data together. And there's a bunch of different components that you need for this project to work which are pretty hard to satisfy. And so a few of these components- or, the first of them, and so I sort of alluded to this earlier, is you need some sort of exogenous variation in the likelihood of having a family member who's incarcerated. Or putting this a different way, we want groups of children who look very similar to each other, where the only difference between them is that one of these sets of children is more likely to have an incarcerated family member, so either a parent or sibling, in our study than the other group. And so a common way of doing this- and so I think so some previous podcast guests have talked about this, is using a random assignment of cases to judges where what happens is that when after arraignment, defendants are randomly assigned to some judges where some judges are stricter than others. And so what this means is that children whose parents are assigned to stricter judges versus less less strict judges are going to be more likely to have an incarcerated parent. And so these children who look very similar prior to this, because it's a random assignment of cases to judges, and so I can I can talk more about that in a little bit to to go through the go through the strategy. But this is going to be providing some source of exogenous variation. And so sort of the first thing that you're going to want is you're going to want a case in which cases are randomly assigned to judges. This is the case in a lot of jurisdictions, but not the case in all.

 

Jeff [00:13:11] The second thing that you're going to need is you're going to need a way of linking children to family members and court data. And so this is actually a lot harder than we had thought going in. So we're going to be using birth certificate data to match children to their parents. That's actually tends to be pretty tricky to get. And so I know of one person who is working on this who put together a lot of really nice data but wasn't able to get the birth certificate data to make it work. And so I think there's probably a number of other cases in which in which that's happened.

 

Jeff [00:13:39] The- sort of the third thing that you need is that we're interested in long run effects on children. And so this means that you need data on what's happening to children over a really long period of time. So first, this means you need birth certificate data if that's what you're gonna use, going back probably 40 or 50 years. I need courts data going back probably at least 30 years, because you need the there to be time for the parents to get incarcerated while their while their children are still in the home. And then you need enough time for the children to get old enough that we can start observing longer term outcomes for them, such as things like their likelihood of engaging in criminal activity as adults, potentially educational outcomes, long run socioeconomic outcomes. It just takes a while. And so you need data going back really far.

 

Jeff [00:14:20] And then finally, you just need a lot a lot a lot of data to have statistical power to say anything interesting. And so in our regressions, we're typically going to have a sample size of over 100,000 children. And honestly, for some of our outcomes, we would have loved to have more than that. And so putting all those different pieces together takes a really long time to do, because really you have to wrangle data from a bunch of different sources and then you have to put all the data together, clean the data. And so it just takes a really long time to do. And if even one of the one of the little components on this process doesn't work out, then you're kind of out of luck. You're you're not going to be able to do the full project. So I think both the data and then also finding a place in which there's some source of exogenous variation, sort of those two things together are really the I think the main reason that it's taken a while for these projects to come out.

 

Jennifer [00:15:10] So your new paper is titled "The Effects of Parental and Sibling Incarceration: Evidence from Ohio." As you mentioned, it's coauthored with Sam Norris and Matt Pecenco. So tell us about this policy experiment that you're exploiting in Ohio. You have data from three counties. You found this natural experiment that lets you measure the causal effect of incarceration. So tell us more about this context.

 

Jeff [00:15:33] Sure. So the approach we're using is one that's very popular in the crime literature. So we're gonna be using random assignment of cases to judges. And so Megan Stevenson did a really good job explaining the strategy during the episode interviewing her, so I'll probably steal a little bit from her in trying to come up with good examples to explain it. But in effect, what we're gonna do or what we're going to want to do is we're trying to cleanly figure out what's the causal effect of parental incarceration. And so here what we really want is we want to see- we want two sets of children who look pretty identical to one another in in all of their characteristics, but where one set of them is going to have incarcerated parents and the other one is not. And so the ideal way of doing this would just be a random experiment, so where defendants are randomly assigned to be incarcerated or to not be incarcerated. And so this will produce random variation whether their children have incarcerated family members. And so the easiest way to imagine this is a judge who's, you know, completely abdicated their duty and they just flip a coin for every defendant who comes in. If the coin lands heads, they incarcerate them. If the coin lands tails, they choose not to incarcerate them. So what that means is that let's say we had a pool of about 2,000 defendants where all these defendants have children. These 2,000 defendants are going in front of this coin flip judge. What this means is that half of the children are going to be randomly assigned to be incarcerated or sorry, half of the children's parents are going to be randomly assigned to be incarcerated. Half of them are going to be randomly assigned to to not be incarcerated. At the start the 1,000 who have incarcerated parents are going to look really, really similar to the 1,000 who don't have incarcerated parents because really the only difference between them is just this flip of the coin. And so what this means is if 10 years down the road, we collect data on the on the children of all these defendants and we see that, say, children whose parents are incarcerated on average have incomes that are a $1,000 lower than children whose parents were not incarcerated, we're going to be able to say pretty confidently this is due to the incarceration, because prior to these children's parents coming into court and being random- in this case, randomly assigned to be incarcerated or not by this judge, these two groups looked very similar. If we see that a big difference between them emerges later, like a $1,000 difference in earnings, we say this is- we think this is due to incarceration.

 

Jeff [00:17:46] And so obviously a coin flip seems like a really unjust system. That's that's not something that we'd want. But our criminal justice system actually does do something kind of like that. And so we're gonna be taking advantage of that experiment where this is going to be based on random assignment of cases to the judge office, where in many jurisdictions after arraignment, so after charges are read, defendants are randomly assigned a particular judge. But judges tend to differ pretty substantially in their severity levels, which is going to be sort of similar to a coin flip. Where in our data in Ohio- so the most lenient judge is going to incarcerate around a quarter of defendants. The strictest judge is going to incarcerate around half of defendants. And so this is a lot like the coin flip where here it's not one judge flipping a coin. It's that your your name is being randomly assigned to a particular judge. And so if you're a lucky defendant, you might get assigned to the lenient judge who only incarcerates around a quarter of defendants. If you're an unlucky defendant, you're getting the bad flip of the coin. You're getting assigned to a judge who incarcerates more like half of defendants.

 

Jeff [00:18:49] And so to sort of to continue with the example is so imagine that there's these 2,000 defendants and imagine you can split them into three different groups. And so let's say that a quarter of them committed really serious crimes. So think of things like murder, aggravated assaults, crimes of a sexual assault, these sort of crimes. A quarter of them committed more medium severity crimes, so think maybe somewhat more serious drug related crimes or something like that. And then half of them committed very low severity crimes. So think of things like, you know, fail- failure to heel your dog in public or something like that. Something where this this isn't this isn't such a serious crime. And so if these two judges are- if these 2,000 defendants are being randomly split between two judges, let's say. One judge who's incarcerating half the time, so that's the strict judge, one judge who incarcerates a quarter of the time, so the more the more lenient judge, let's think about what's happening to each of these different groups of defendants. So let's start off with the 500 defendants who committed really serious crimes. So 250 of them are going to go to this really lenient judge, 250 of them are gonna go to the strict judge. And so the lenient judge's incarcerating only about a quarter of the people who come before them. Presumably, they're going to be incarcerated the quarter who are committing these really serious crimes. And so these 250 who go to lenient judge are going to be incarcerated and the 250 who go to the strict judge are also probably going to be incarcerated. So all these defendants who are committing the more serious crimes are probably going to be incarcerated regardless of the judge that they're assigned to. Similarly, if you look at the 1,000 who committed the crimes that are that are not so serious. So half of them are going to the lenient judge, half are going to the strict judge. In both cases, the strict judge's only incarcerating about half of people, presumably they're not incarcerating the people who committed the least of the least severe crimes. And so all the people who are not committing the more severe crimes are probably not going to be incarcerated.

 

Jeff [00:20:46] What's going to happen is for the people who are committing these more medium severity crimes, there's going to be about 250 of them assigned to the strict judge and 250 to the lenient. So these are the ones where the judge to whom you're assigned is probably going to matter. Where, if you're assigned to the strict judge, this judge tends to incarcerate these types of offenses where the lenient judge tends not to. And so what this is going to mean is that the children of these defendants are going to differ in the likelihood of having a parent who's incarcerated, where those assigned to the strict judge are going to have- or be of be much more likely to have an incarcerated parent. Those assigned to the lenient judge are going to be much less likely to have an incarcerated parent. And so what that means is that these two groups of defendants' children, look really similar at the start before their parents are randomly assigned to the judge. If later down the road, we collect information from them. So let's say that we see that the children whose- the 1,000 children whose parents were assigned to the strict judge have earnings that are 250 dollars lower than the 1,000 defendants whose children were assigned to the the less severe judge, we are going to say this is attributable to incarceration. And given that we know that this difference in incarceration rates between these two judges is about a quarter of the population, we can really scale this to say this is going to amount to about a loss of a thousand dollars in earnings as a function of having an incarcerated parent.

 

Jeff [00:21:58] And so what we're going to do is something slightly more complicated than that. So we're going to be using all the judges in our data and the probabilities of incarceration on a scale between a quarter and a half. So we're going to be using judge severity as an instrumental variable for likelihood of having an incarcerated parent. That's sort of the intuition for the strategy that we're going to be using.

 

Jennifer [00:22:16] Whenever we read these these judge randomization papers in class, my students are always horrified that there's so much variation across judges and the likelihood that they'll incarcerate people. And of course, that's not not ideal for justice, but is certainly good for research. And so so as you've just explained, you know, this this type of identification strategy is going to measure the effect of incarceration for those on the margins, for those for whom being assigned to a different judge might actually lead to a different outcome. So for these these medium severity offenders, in your example, who are they in in your context, who should we think of as the relevant population here that your analysis is going to be telling us about?

 

Jeff [00:22:56] Definitely. So they're actually going to be in a lot of ways - the defendants who we see who are the marginal defendants are the ones who are being affected by the judges to whom they're assigned - in a lot of ways, they actually look pretty similar to the overall population in terms of demographic characteristics. Where, for example, they're no more they're no more or less likely to be parents than general population- than the overall population. They're more no more or less likely to to be black versus white. They are more likely to have committed crimes that are related to drug offenses. They're much less likely to have committed very serious crimes. So, for example, a crime such as murder. So, for that type of that type of case, really, the judge to whom you're assigned, that's probably not going to that's not going to affect whether or not you're incarcerated- if you're found guilty, you are almost certainly going to be incarcerated. Though, in fact, in our case, murder cases in the state of Ohio are not randomly assigned. So that's a bit of a bad example. But think a very serious aggravated assault or something like that. And then again, the cases which are much, much lighter, so think maybe a very simple marijuana possession or something like that. These are the cases in which individuals are much less likely to be incarcerated and they're not really going to be affected by the judge to whom they're assigned. So the types of cases you should be thinking about are sort of medium severity, drug drug cases, less severe cases of assault, so simple assault or something like that. These are more the sort of cases that we'd expect to be affected by the judge to whom someone is someone is assigned.

 

Jennifer [00:24:23] OK, and then as we're thinking about the treatment of incarceration, what should we think of as the counterfactual here for the kids? So if someone's mom or dad is locked up, who do they typically live with instead? Do- can you see that in any way?

 

Jeff [00:24:37] Yes. So it's going to depend a lot on the identity of the- whether it's a mother or father. So when it's the case of the father- so around 85 percent of the time, the children are going to live with their mother. And so in many of these cases, the children already live with their mother, and the father may not be currently residing with them. So that's not something that we can see in our data, but this is something that in some other survey data, people have shown. And then in the rest of the cases, they're mostly going to be living with grandparents or other relatives. This may be because the children are removed from the father's care and taken to live with the grandparents or other relatives or in many cases - and actually probably a little bit more like a majority of cases - they're already living with those relatives. In the case of mothers, so here about 40-45 percent of the time, they're going to be living with their grandparents. And about, I think 30 to 35 percent of the time, it's going to be the father. Around 20 percent of the time, it's going to be other relatives. But it's actually not very frequently going to be foster care. So only about 11 percent of the children with incarcerated mothers are going to be- are going to be in foster care. For children with incarcerated fathers, it's gonna be an even lower figure. So it's going to be a lot closer to two percent. I think my percentages probably didn't all end up all add up there, but that's roughly the percentages that we're talking about, where most of the time for mothers, the children will be living with grandparents or with other family members, but very rarely in foster care.

 

Jennifer [00:26:03] OK. And I think that's really important because I think most people probably have in mind that the kids are going to be thrown into the foster care system. And if in practice, they actually wind up potentially living with more stable family members, that could help explain why sometimes there are beneficial effects. OK. So you you talked a little bit about the cool data you're using. You have a whole bunch of administrative datasets from Ohio. And I think this paper provides a really great example of being creative and resourceful and finding ways to quantify things that might seem difficult to measure at the outset. So tell us about all the cool data you've been able to gather and link together for the study.

 

Jeff [00:26:40] Thanks, yeah. So the data that we have is, as you said before, it's from the the three largest counties in Ohio. So this is Cuyahoga County, Franklin County and Hamilton County. Folks probably know this better as the counties surrounding the cities of Cleveland, Columbus, and Cincinnati. And so together, these different counties, they have a combined population of about three and a half million. And something that's nice, actually, about working in Ohio- so this isn't something that we realized going in. This is something that we realized later and we're sort of we're sort of glad that it ended up working out in Ohio. So I should I should mention here as well that this was not the first place that we tried to do this, this this project. We had to look at a lot of different places, trying to find a setting in which we were able to get all this data and put it all together. One of the things that's nice about Ohio is it's actually very similar to the rest of the US when it comes to crime and criminal activity. So the crime rate in Ohio annually is about 4,000 crimes for 100,000 people. That compares very closely to the overall US average, which is a little bit under 4,000. So it's very very close. Recidivism rates, incarceration rates are also really, really close to the overall US average. And so we think that that this this the results that we get from Ohio are hopefully going to be able to generalize outside of outside of the state of Ohio. And I'll probably come- we can come back and talk more about the generalization and heterogeneity at the end, because I think especially when thinking about many of the other really nice papers that have come out over the past year or so using a very similar strategy to ours, I think it's gonna be important to think about why do we perhaps see somewhat different results across different settings?

 

Jeff [00:28:09] And so our main result or main approach is going to be to use administrative data. So this is going to be data collected by by government agencies where basically what we'll do is we'll go around and try to form a data usage agreement where they'll they'll let us use their data, sometimes with some some restrictions on that in terms of some fields that are anonymized or things like that. And so the different data that we're gonna be using is, first of all, really the basis for this is going to be courts data. And so we're gonna have court records for the- what amounts to felony courts and misdemeanor courts in all three of these counties for adults. And so we can see anyone who appears in these courts as a defendant going back in in most of our counties to around the early 90s. What we're then going to do is we're going to take about 50 years worth of birth certificate data. Then we're going to use that to match parents to children. So if we see someone appears as a defendant in court, we're gonna then look to see do they appear as the mother or father in birth certificate data? And then we're also gonna be able to match children to their siblings in a similar way where we'll look for people who are criminal defendants in in the courts. I'll look to look- match them to their birth certificates to figure out who their mother and their father are and then match them to their siblings, through people who have either a common mother or common a common father. And so that's really going to be- those two sources of data are really going to be the key things that we use. In terms of trying to get this- these measures of whether a child has an incarcerated parent and then also for getting this exogenous variation and likelihood of having an incarcerated parent, we're gonna use court records on on the judges and use this to figure out who's a severe judge, who's a who's a less severe judge.

 

Jeff [00:29:43] In terms of outcomes data, so we're going to be using the court data for one. So we're going to be looking to see, you know, 20 years down the line, what's the likelihood that a child who has an incarcerated parent commits a crime or is accused of a crime themselves and then also potentially is incarcerated for for the crime that they're accused of? We're also gonna have data from juvenile courts in in the city of Cleveland. And so this is going to allow us to measure not just adult criminal activity, but also juvenile criminal activity. We'll also have data on academic performance, so of school records from Cleveland public schools. So here we'll be able to measure things like children's test scores on on large tests, like the No Child Left Behind test. I'll also be able to measure grade repetition. And so we'll have a little bit less of the school's data. So we'll have eight years worth of schools data. For the other data, we usually have more on order of 20 or 30 years, but still enough to, I think, say something interesting. We also have the birth certificate data. We can use to measure teen pregnancy. So after doing this matching, we then look to see the children whose parents are criminal defendants, do we see that they appear as as parents themselves as as teenagers?

 

Jeff [00:30:51] And then finally, the last thing that we're gonna do is we're gonna be trying to measure what's the neighborhood in which children live as adults and use that as a measure of their adult socioeconomic status. Where the idea of what we're going to be doing here is we're gonna figure out what's the address that a child lives in as an adult. We're then going to match this to American Community Census- sorry, American Community Survey data on the wealth of in a particular census block group. So this is a very low level of geographic aggregation. So having about, you know, 1,000 to 3,000 people in it. This measures the poverty level of people living within that census block group. And so we'll be able to see does the child's, after they grow up after they're an adult, are they more likely to live in a wealthier or poorer neighborhood? This potentially one is just serving as a proxy measure for how wealthy or how poor they are, as well as there's a lot of nice recent research looking at neighborhood effects. And so potentially, the neighborhood in which the child lives in as an adult is gonna have some important ramifications for for them in terms of ability to find work, for example, as well as for their children in terms of their children's later life outcomes. I think there's- so there's one other data set that we also use, so we'll be using data from evictions court. So we'll be using that to try to measure after a child's parent is incarcerated, what effect does this have on their economic outcomes by looking at what's the likelihood that their their family is evicted? And there's a few other- I think that covers the main datasets that we use. There's a there's a large number, and so hopefully listeners can appreciate this is why it took us a really long time to put this together and why all of these projects that have been working on- have been have been working on this have just taken a really long time to to to get off the ground.

 

Jennifer [00:32:27] Yeah, and just in case anyone's curious: you match all of these, I think, using name and birth date, was that right?

 

Jeff [00:32:33] Yeah, that's right.

 

Jennifer [00:32:33] Because they're not datasets where you have, like, security numbers in all of them. It's usually not the case in in crime data in particular, which makes it even harder.

 

Jeff [00:32:41] Yep.

 

Jennifer [00:32:42] OK, cool. So let's dive into the main results. So what do you find are the effects of parental incarceration on their kids' criminal activity?

 

Jeff [00:32:50] Sure. So, I should be very upfront with- about these as well. So the results that we got were very surprising to us. I think going into this project, we all had this idea that this- the effects of parental incarceration in particular, particular something like maternal incarceration, were gonna be pretty negative. And so when we saw these results come out in Stata, needless to say, we were a little bit surprised by what we found. Which is that children whose parents are incarcerated - so using this random assignment of cases to judges to look at what's the causal effect of having an incarcerated parent - are actually less likely to commit crime themselves as adults or as as juveniles and less likely become incarcerated themselves as either as adults and as juveniles. What we see is that the child's likelihood of ever being incarcerated goes down by about 3.2 percentage points, which is a pretty large effect. And these effects are concentrated among children who live in the poorest neighborhoods.

 

Jeff [00:33:43] And so, as I said, this is this is not something we that we expected to see. And there's there's a number of different mechanisms that potentially could explain this. But in general, what it looks like is that having incarcerated parents seems to be having, at least when it comes to criminal activity, somewhat beneficial outcomes on the child's later later life outcomes in terms of committing crime. The thing that I think is nice in looking at both juvenile criminal activity and adult criminal activity, is we might imagine that maybe there's short- different short run and long run effects, where maybe the effect of having an incarcerated parent maybe leads to some emotional dislocation or trauma such that children may be more likely to commit crimes in the short run. But perhaps for a variety of reasons, this may make them less likely to commit crimes in the long run. That's not what we see. What we really see is there's a short run reduction in crime and juvenile crime and also a longer run reduction in adult crime. The adult crime results are actually a little bit less strong than the juvenile results, particularly when you look at the subsample of children who are born in poor neighborhoods. The effects are pretty strong that there does seem to be a reduction in criminal activity.

 

Jennifer [00:34:47] And you mentioned potentially differential effects by whether it's the mother or father incarcerated. Do you find differences there?

 

Jeff [00:34:55] So, there are slight differences. They're not statistically significant. And so I hesitate a little bit to to say whether whether whether one is is more or less strong, because we can't reject that that the two are the same. But if you're just strictly to look at the point estimates, it does look like it's a little bit stronger of an effect for maternal incarceration versus paternal. But again, this is one of those things where, you know, we have 140,000 observations here. We'd love to have, you know, 500,000 observations to really be able to tell the difference between maternal and paternal. Even though suggestively it looks like it's maybe a little stronger for maternal.

 

Jennifer [00:35:33] Yeah, I think that would be consistent with at least anecdotal evidence that, for a lot of judges anyway, they need to feel like the woman standing in front of them is in particularly bad shape in order to lock her up if she's a mom she's a mom and might not be as reluctant if if it's a dad. And so the marginal mother might just be higher risk than the marginal dad. Do you see any differences by the gender of the kids in terms of outcomes?

 

Jeff [00:35:58] Yeah. So here's- it's again a case where we can say there's some suggestive effects, but we can't reject that the estimates that we get are statistically not different from one another. So we do see larger effects in terms of the point estimates for boys. So, for example, for for juvenile incarceration, I think we see a reduction of about 3.9 percentage points, whereas for for for women or for girls, we see a reduction of, I think, 1.8 percentage points. But this is really just coming- the main reason for this is just boys are much more likely to be accused of criminal activity and be incarcerated than than girls. So I think a lot of that is just just coming through that, it's just that girls are just much less likely to commit crime in general.

 

Jennifer [00:36:40] OK. And then you you do have all these other outcome measures, so you have teen parenthood, long run socioeconomic status, evictions. What do you find there?

 

Jeff [00:36:48] Yes. So we get generally somewhat consistent results, though it depends a little bit for the- across the different different outcomes. So we don't see any statistically significant effects on test scores. So here we have a positive point estimate, but nowhere near statistical significance. I don't- I think the p-value is probably around like 0.2 or 0.3 or something like that. And so we don't find any statistically significant effect. Our standard errors are moderately sized, so we can rule out medium size negative effects on children, which is given that our prior here - the thing that we thought and I think most people when we first talk about this project thought - was that this was going to have pretty negative effects on children. We can rule out, you know, moderate size negative effects on a child's test scores, child academic outcomes. For teen pregnancy, so again, here, we're not gonna be able to say that the the effect that- we got a null effect, so we can't rule out that the effect is zero. Here, our standard errors are gonna be a little bit bigger. And so we're going to be a little bit less confident in saying that there is or isn't an effect on teen pregnancy. This is also just a more subtle outcome to try to to try to detect. So, we you know we can rule out relatively large effects on teen pregnancy, either positive or negative. But we can't rule out a zero effect.

 

Jeff [00:38:07] But really, the one that we have the most power to look at - and the one again, this was this was pretty surprising to us, we didn't really expect to see effects of this magnitude - was on long run socioeconomic status. And so as I said before, what we look at is we look at adult residential address. We're geocoding this to see what's the average poverty level of the neighborhood in which in which children live in as adults. And generally, what we find is that children whose parents are incarcerated actually tend to live in in wealthier neighborhoods as adults where - for whatever reason, and this is something that we we can say with confidence - is that this effect seems to be stronger for girls than it is for boys. And so we- it may just be the case that maybe the there's some other recent evidence looking at this at intergenerational mobility for girls versus boys. It may just be that there's more elasticity for girls in terms of intergenerational mobility than there is for boys. So maybe that's why the effect is stronger. We don't have a very good- we don't have a very good understanding of that right now. That's that's one of the things if we had, you know, 500,000 or a million observations, maybe we'd be able to dig into a little bit more. For now, really what we can say is there does seem to be this positive effect on socioeconomic status, mostly driven by girls.

 

Jeff [00:39:16] But broadly, the way we sort of interpret these findings is that in general, it seems like on net that these are relatively positive. Of course, though, there's gonna be a lot of heterogeneity here. And so, as I said at the beginning, there's almost certainly going to be some cases in which some children are going to be hurt, almost certainly some cases in which children are going to be harmed. The effect that we're gonna be capturing is the net effect of those two different effects. We find that the net effect is positive, but that certainly doesn't mean that there are some children who are who are going to be harmed by this. And that's obviously something that's going to be pretty important to think about, especially when thinking about what are the policy implications of this study or something like that. I just want to be very clear we're not saying that incarceration of parents is good for everyone. What we're saying is that on net for, I think as you very nicely pointed out, for this marginal population, for these defendants of- for this population of defendants who are who are marginal in terms of incarceration or not, it does seem to be that the effects are, on net at least, relatively positive.

 

Jennifer [00:40:12] And then so that's all for parental incarceration and you're also able to link the folks who are incarcerated to their siblings. And so you can measure the effects of sibling incarceration. What do you find there?

 

Jeff [00:40:25] Yeah, so for sibling incarceration- so here, we're gonna be a little bit limited by sample size and the outcomes we look at. And so for here, we solely focus on the outcome of adult criminal activity. So what is the effect of having an incarcerated sibling on the likelihood that the child is is either accused of a crime as an adult or is incarcerated as an adult? What we find here is is very, quite large effects, quite large reductions in adult criminal activity as a function of having an incarcerated sibling. And so the reduction that we find is around 6.7 percentage points, which is a really, really large effect. This is larger for for boys than for girls. And here we can- we're close to being able to say this is a larger effect for boys than for girls. But again, this is just because boys just commit crimes at a much higher rate than than than women do.

 

Jeff [00:41:12] And so there's there's actually one other recent paper I should flag here that looked at this in Norway using the same identification strategy that we're using- using random assignment of cases to judges. And they find something very similar to what we do, where incarceration of siblings greatly reduces the likelihood of the the other sibling being arrested for a crime. The effects that we find are actually- so our effects are, we think, are pretty large. The effects they find there are actually even larger than the ones that we find. And in general, we do a little bit of work looking into mechanisms. It looks like a large part of the reason for this is, is what's happening when, particularly in the case of removal of siblings, is this maybe removing a criminogenic influence or an influence that makes the child potentially more likely to engage in criminal activity. This could potentially be because the siblings are introducing one another to criminal peers. Maybe the siblings are committing crimes together. There may be some some sort of peer spillover mechanisms that are that are going on there. So that really seems to be the story there. It's a little bit of a different story than the parental incarceration story. But nonetheless, we find somewhat similar results, I think, for a slightly different set of reasons. And I think really the sort of more like a peer spillover effect is the reason we spend so much larger effects for siblings than we do for parents.

 

Jennifer [00:42:22] Yeah, so let's talk more about all the nice tests you do to try to tease apart mechanisms. I think that's a really nice contribution in this paper in addition to, you know, the incredible data undertaking here. Yeah, so you test a bunch of different channels through which parental incarceration and sibling incarceration could be affecting these outcomes. So tell us about those different tests that you run and what you find.

 

Jeff [00:42:44] Sure. So the- there's a bunch of different tests that we run. And so for, as you said, like for a phenomenon this complicated for removing a parent from the life of a child, there's a lot of different things that are going to be going on here. And so the story that we're have is- the story that we find is certainly not going to be a simple story where there's one mechanism that's driving anything, everything. It really seems to be like more of more of a multicausal story. And so the first thing that we wanted to look at is as economists, we think that, you know, economics matters, resources available to the household matters. And so something that had been a strong prior of ours was that when you remove a parent from the from the household, from from the child, that may be removing some economic resources that potentially could be beneficial for the child. So one of the first tests that we wanted to do is look to see, is this actually the case? Is it the case that when you remove a parent from the household based on you incarcerate the parent, is it the case that this, at least in ways that we can measure, worsens the economic status of the child?

 

Jeff [00:43:46] And so the two ways that we do this is the first, as I mentioned a little bit before, is we look at evictions court data. And so we look to see- so for for children, so we look to see if one parent is incarcerated, do we see that the other parent, which is predominately where the child is going to be living, is the other parent more likely to show up in evictions court records as having an eviction notice served against them? And so we find no difference in likelihood of having an eviction notice served against them. So this is a measure of a very extreme type of financial distress. Another thing that we do is we're gonna use the data on resi- on residential address. And so we look to see when one parent is incarcerated, do we see that this causes the other parent to be living in a neighborhood that's lower socioeconomic status than than where before? And so we again, we don't see an effect on- of having- so let's say we're taking we're taking Parent A, we don't see an effect of Parent A being incarcerated on the socioeconomic status of the neighborhood that Parent B is living in, where this is a proxy measure of the neighborhood that the child is living in since typically they're going to be living with with Parent B. And so that doesn't- so our prior was this was gonna be a strong effect. It doesn't seem to be the case. I think the reason for this is something you alluded to earlier, which is that in many cases, the parent who's on the margin of incarceration, they may be only potentially a marginal source of income for the household. And in some cases, they may actually be may be draining some economic resources for the child- from the household. Particularly in cases where the parents have some sort of substance abuse issue or something of that nature. So this may be the reason that we don't see very- we don't really see much in the way of economic effects of this.

 

Jeff [00:45:28] A second mechanism that we wanted to look at is you know, maybe what's happening here is that there's rehabilitation. Where what's happening is the parent is incarcerated. This incarceration leads to rehabilitation. So there's a really nice paper Bhuller et al. in Norway that finds that incarceration of of individuals actually can improve their later life outcomes by by rehabilitating them. And so we look at rehabilitation to see after- if we take the parents where some of them are incarcerated, some of them are not using this random assignment of cases to judges as an instrument for incarceration. What we find is that being incarcerated initially reduces the amount of crime that the parents are committing. This is really just due to the fact that the parents that the parents are incarcerated are incapacitated. And so they're not able to commit crimes. But once they're released from from either jail or prison, we say that they don't commit crimes really at any higher or lower rates than the parents who are not incarcerated. And so it doesn't seem like there's really that much in the way of rehabilitation of the parent who is incarcerated.

 

Jeff [00:46:24] Something else we look at, which is a very similar test, is to say OK maybe the parent who's incarcerated is not rehabilitated, but maybe the other parent changes their behavior or changes their activity as a result of the other parent being incarcerated. So Parent A is incarcerated, maybe Parent B changes what they're doing. So maybe they, say, become less likely to engage in criminal activity because they know they're really the only parent who's still there and they need to be at home to take care of the child. They can't afford the risk of engaging in criminal activity. And so we do see some evidence of this where the parents who- the nonincarcerated parent is a little bit less likely to be incarcerated themselves. So that that seems like some evidence of this, though the effects are relatively small.

 

Jeff [00:47:07] Places where we think there is potentially more going on here is two other mechanisms: a deterrence mechanism and then a removal of a criminogenic influence mechanism. And so the idea behind the deterrence mechanism is that children, through experiencing the criminal justice system, through the eyes of their parents, this may make them less likely to commit crimes themselves later in life. Or, for example, if they see the negative, potentially emotional impact that being separated from their parents has on them, they may say I'm less likely to engage in criminal activity because I don't want to do this to my family members or or my children. And so this could potentially reduce their likelihood of engaging in criminal activity.

 

Jeff [00:47:43] One interesting mechanism that, again, sort of ran against my priors going into this, though it's one that's consistent with with something from a nice ethnographic work by Megan Comfort called "Doing Time Together," which follows the romantic partners of incarcerated men in California. One thing that we might imagine that happens is that after a parent is incarcerated, this may lead to relationship dissolution where if Parent A is incarcerated, maybe this makes it less likely that they're going to continue to be in a relationship with Parent B and make- or through that, may continue- may be less likely to have a relationship with their children. On the other hand, this isn't necessarily obvious what's going to happen. Where it may be the case that by being incarcerated, this may reduce the parent's outside options or it may cause them to reprioritize in such a way that causes them to prioritize the relationships that they had before they were incarcerated. We actually find some evidence of that, where what we do is we look at the parents who are incarcerated, so let's say Parent A is incarcerated, look to see what is the likelihood that they have a child with Parent B, the parent with whom they already had a child prior to being incarcerated versus having a child with some other with some other parent. So there's Parent C out there who they haven't had another child with. And what we find is that people who are incarcerated for this quasi random reason of random assignment of cases to judges are actually more likely to have children with the mothers of their children prior to their incarceration and less likely to have children with new parents or with with other with other mothers in this case. We only see this for fathers. So fathers are less likely to have parents with with new mothers, more likely to have children with with the mothers with whom they previously had a relationship. And so this maybe means that they're more likely to be involved with the mothers afterwards, and potentially more likely to be involved with the with the child. So this is potentially one mechanism through which there might be some some positive effects.

 

Jeff [00:49:38] Again, though, we always have to think about what's the heterogeneity here where there's certainly probably some population of parents for whom that's true. Other ones for which relationship dissolution may be more true. But it may be the case that those parents who recommit to the relationship, maybe those are the ones who are the most beneficial relationship in their child's life going forward. And so all that to say is that there's a lot of really complicated things that are going on here. We have some suggestive evidence on some mechanisms. There's some I think that we can kind of rule out, other ones that we don't have as good of evidence on. But it's really just a complicated story that's that's going on here and a lot of different things are playing into the effects that we find.

 

Jennifer [00:50:13] So as you mentioned, there there have been a bunch of new papers on this topic released in the past year or so. We went from having almost no well identified research as, you know, as far as economists would consider well identified on this issue to having five or six studies. I think you're right about that. I still haven't had a chance to sit down and read them all side by side, but I suspect that you have since you've written about this. So I want to pick your brain about them. So tell us about those other studies and what else we've learned about the spillover and intergenerational effects of incarceration since you and your coauthors first started working on this paper.

 

Jeff [00:50:46] Definitely. So I think it's, I should say, up front, I think it's great for science that we have all these different results, especially since a lot of them are coming from different contexts. And as I'll say in a second, I get slightly different results. And so I think having all of these really emphasizes potential heterogeneity. And I think that's just really important in thinking about a problem that's as complicated as parental or sibling incarceration. And so just to go through the list of of pieces of evidence that we have. So I think I already talked about the Billings paper in North Carolina. And so this is looking at short run effects of parental incarceration. On there, what he finds is relatively positive effects on things like behavioral outcomes, where children immediately after their parents are incarcerated, they're actually less likely to have behavioral problems in school than the children who- yeah have behavioral problems in school. That's a slightly different identification design, so it's using student fixed effects, so really comparing the child's when their their parent is- after they're arrested or after they're incarcerated to when the parent was not arrested or not incarcerated. So, a slightly different strategy, but I think informative nonetheless.

 

Jeff [00:51:45] There's going to be a number of papers that are very similar to ours in using the same identification strategy - so these are going to be using random assignment of cases to judges. And so these are gonna be coming from a bunch of different contexts. So there's one paper in Norway, one paper in Finland, one paper in Sweden, one one paper in Colombia. And so all these are going to be using random assignment of cases to judges across those different jurisdictions. And what they find is actually gonna be fairly different from one another. So starting off with the Norway and the Finland paper- so the paper in Norway- so they're- what they find is they find a null effect. However what the paper states, and so I think this is a direct quote or a roughly close to that, is that the IV estimates are too imprecise to be informative. And so really, what's, I think an issue that they run into - and this is something that we experience as well and I think a lot of IV IV papers do - is just the sample sizes in Norway just were not large enough to be able to say something super convincing about whether there's a positive or negative effect. But in general, the effect that they find is a null. The Finland paper, again, is going to have a bit of an imprecision, some issues. They get some more some more compelling results on what's the direct effect of incarceration for parents. For the other things that they look at, they find more null or perhaps slightly negative effects of having an incarcerated parent on the child. That paper is- they're still working on that. I think it's really exciting. They've got some really nice outcomes that they're looking at there, so I'm excited to see the final draft of that.

 

Jeff [00:53:10] The two other papers, one is in Sweden and one is in Colombia, so I'll keep going with the Scandinavian countries and and start off with Sweden. And so what this paper is going to find is they're going to look at a bunch of different outcomes. And so they're going to have a bunch of different things that they can look at. I have to say I'm very jealous of all the all of the Scandinavian data. I think me and many other people working in the US, when we see the many amazing things that they can look at, it it just makes it it makes me wish that we had the same thing in the US. But some of the things that they find is- so one thing that they look at is on teen pregnancy. And so they find very, very substantial increases in teen pregnancy as a result of having an incarcerated parent. And so I think the effect size is it's an increase of something like 176 percent over the dependent variable. And so this is a really, really large effect. They also find some increases in juvenile criminal activity - again again, fairly large increases - as a function of having an incarcerated parent, a lower likelihood of being employed at age at age 20. So these are going to be a little bit more short run effects than ours since they're going to be looking at mostly prior to age age 20. But again, I think it's pretty easy to say that these are going to be negative effects and the negative effects that they find there are almost certainly going to extend to the later time period.

 

Jeff [00:54:25] And then finally, the paper in Colombia looks at looks at schooling. And so there, the primary outcome is going to be looking at years of schooling. And what she finds is that having an incarcerated parent, particularly incarcerated mother, leads to relatively positive effects where children whose whose parents are incarcerated achieve more schooling than children whose parents are not incarcerated. And so this is a- maybe at first seems like a hodgepodge of different results, and indeed it is hard to say why are different people getting different, different estimates. But I think there's a few things that we can we can learn from this. I think it's great that we do have all these estimates in order to learn from. And so I think one of the primary things that that's going on here is that just there's very different populations potentially of people who are being incarcerated across all these different contexts. And so, for example in Colombia Colombia, the defendants there who are on the margin of incarceration, it looks like are probably committing slightly more serious crimes than the ones that we see in the US. And so potentially, if defendants are committing somewhat more serious crimes, then one reason why these effects may be so positive is potentially because it's removing defendants who are potentially having a more negative influence on the children through unstable home environment or the like. In general, though, I'd say I think that our results and hers are very similar. And so I think in general, I think there- our two papers are pretty broadly consistent. I think a reason for that is just that the criminal justice system in Colombia is much more similar to the US potentially than than in Scandinavian countries.

 

Jeff [00:56:01] And so in Sweden potentially - though I actually I don't know the answer to this, I should I think though the draft- the earlier- the previous draft that I looked at, I don't think had that much evidence on this, so I should go look to the most updated draft. But my guess is some things that may be going on there is one that the defendants may just be committing on average, less serious crimes. So there may be reasons that have to do with that particular context for which which mean that removal of the of the parent is potentially having a more negative outcome. There may also be more of a deterrence effect in Colombia and the US, where Scandinavian prisons are sort of known for being less severe, shall we say, than than US prisons or prisons in Colombia. And so maybe there's a stronger deterrence effect of seeing a parent who's incarcerated in the US or Colombian context.

 

Jeff [00:56:49] In general, though, what I'd say is I think we don't have- I think we can speculate. And so everything that I just said is complete speculation. But in general, I think all the estimates that we get are, you know, very relevant for our contexts. I think those are good papers. I really I don't have I don't have quibbles with the papers. I think the estimates we all get are telling us something really important about the different contexts that we're looking at. And so I think that our estimates are great and very applicable in the context that we're looking at. I think what this tells us however is there seems like there's a fair amount of heterogeneity across space. And so though I suspect that people may not have that much of an appetite for, you know, coming up with the sixth and the seventh paper on parental incarceration using random assignment of cases to judges, I think it would actually be great to see more evidence on this other contexts and to try to understand more on the nature of this heterogeneity. But in general, I think, you know, there's there's a lot of heterogeneity in this. There's a lot of mechanisms that are going on here. And so I think that that's- it it's maybe not the most satisfactory answer to say I think there's a couple of reasons why this might be true, but we don't know for sure, but that's the best that I have.

 

Jennifer [00:57:50] Such is the nature of research. Yeah, I think it's the same thing, actually, in just measuring the effect of incarceration on the person incarcerated. We get a bunch of different studies, all measuring, you know, measuring the effects in different contexts. They've got different marginal offenders, different treatments, say incarceration means different things in Norway than it does in the United States, and they find different estimates. And I think we're really just at the beginning of trying to sort out why, you know. What's leading to the positive effects in some places and negative effects in other places. And obviously, that's important. We want to design a better system. OK. So what are the policy implications of your paper and the other work in this area? What when policymakers come to you and ask what they should do based on your results, what do you tell them?

 

Jeff [00:58:36] Yes. So the first thing that we that we would try to do in this case is to be very careful, because I think there's a lot of ways in which these results could be applied in ways that we that we don't really think are right. And so we would be very clear that nothing in the paper- in the paper, we don't say what our results are saying is that we should be locking up more parents. That's that's definitely not what we think. That's that's that's not the takeaway that we'd have from this. The way that I would read our results is to say that, for a set of children who are in a very highly vulnerable situation - so these are children whose parents are on the margin of incarceration - we are seeing these positive consequences of this particular policy intervention, which is to say incarceration. This doesn't necessarily mean we should be incarcerating more people. Incarceration is a highly costly policy. There's also probably emotional consequences that we're not going to be able to measure here.

 

Jeff [00:59:19] But what it does say is that this set of children who have parents who are on the margin of incarceration are fairly elastic in their later life outcomes. And so we see that this intervention just, you know, a pretty major one is having a pretty substantial impact on them in ways that are reducing their likelihood of future criminal activity, improving long run socioeconomic status, both things that are really, really difficult to shift around. And so what we think is what this means is that first, this is a population that there should be more policy attention focused on and that the probably probably the best policy intervention is not necessarily going to be incarceration. There may be other potentially cheaper ways of of of having some of the same results, but without the negative consequences that we would imagine from incarceration, both potentially on the child in terms of emotional consequences, as well as other reasons, such as budgetary reasons or all the other negative consequences that we've seen for direct consequences of incarceration.

 

Jeff [01:00:14] I think also for something that's helpful to think about here. And I again, want to be very careful in my language or the way that I speak about this. I think what this really highlights is the heterogeneity here. Where there's certainly some children who are being harmed by having incarcerated parents and some children who are being potentially helped. And so what I think this highlights is that, one, just the importance of the role of the judge where if they are taking this into account in their in their sentencing - and indeed in some cases they are - understanding that there are some cases in which it's going to be more beneficial than others and trying to think about sentencing in those ways. Where we had on net for the people who are marginal, we find these positive effects. But it certainly doesn't mean it's the case that the judges should start incarcerating massively larger numbers of parents in order to to achieve some of some of these ends. But more broadly, I think just in terms of thinking about future research as well, understanding this heterogeneity, I think, is going to be is gonna be really important. And aside from trying to come up with interventions that are beneficial in the lives of children who whose parents are on the margin of incarceration, I think understanding what are the characteristics that make the effects be beneficial in some cases and more detrimental in others is a really important next step for for future research, and potentially may also help us in resolving some of these questions we were just talking about in terms of why do we see different results in Scandinavian countries than we do in the US and Colombia. By the US, I mean, both both our paper and then also the North Carolina paper as well.

 

Jennifer [01:01:49] So that's a great segue into the- my research frontier question. So so that is certainly a research frontier. We we need a lot more research on that. Any other big questions that have come up in the course of this research that you think that you and others working in this area will be thinking about in the years ahead?

 

Jeff [01:02:05] Yeah so I mean, in general, I think just focusing on this population of children whose parents are incarcerated is just a really important and interesting population to focus on. So there's- the estimates we actually have are not very good on what what's the number of children who have incarcerated parents, but that's probably on the order of around five million. And so, you know, five million people in the US who have had parents incarcerated at some point during their childhood, this is a really large population. So I think coming up with interventions that are focused on this particular population are potentially quite interesting. In particular, thinking about different policy interventions, such as changing around visitation policy, changing around the ways that people are assigned to prisons that can be changed around visitation, things like that, thinking about alternatives to incarceration and seeing what is the effect that that that this has on on children. I think it's also I think all these things are potentially really exciting just focused strictly on this on this area of parental and sibling incarceration.

 

Jeff [01:03:01] And then just to reiterate, I think the heterogeneity here is just incredibly important. I think in order to get at that we're going to need more and larger scale administrative data. And so I think that, so this is something we've talked about and I've a few thoughts on ways that that one could potentially do this, beginning a larger scale administrative side to try to get a bit more of this heterogeneity, try to understand that better, I think is potentially really important follow up, follow up to this. And it just more broadly, I think thinking about crime research generally, just as you said, is we do see in the crime literature that there are these really heterogeneous outcomes in terms of, say, the direct effect of incarceration, where there's some studies showing that even even within the US, there's some studies showing, you know, incarceration reduces later criminal activities. So there's a really nice recent paper by by Shem-Tov and Rose, that was a job market paper this year. Then there's also other papers showing that this has pretty negative consequences. It causes people to be more likely to commit crimes once they're released, like Mueller Smith in Houston, which is another great paper. And so I think getting more data like this, that'll enable us to try to understand what is causing this heterogeneity to me just seems like really the first order thing. And so it's something we were thinking about a little bit now in terms of follow up research on this. But I think that to me that that seems like really what's crying out for for better answers in the economics of crime literature.

 

Jennifer [01:04:19] My guest today has been Jeff Weaver from the University of Southern California. Jeff, thanks so much for joining me.

 

Jeff [01:04:24] Thank you so much for having me.

 

Jennifer [01:04:31] You can find links to all the research we discussed today on our website, probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show. Our sound engineer is Caroline Hockenbury, our music is by Werner, and our logo is designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.