Episode 28: Jillian Carr

 
Ep28_Carr_photo.png

Jillian Carr

Jillian Carr is an Assistant Professor of Economics at Purdue University.

Date: April 28, 2020

Bonus segment on Professor Carr’s career path and life as a researcher.

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Professor Carr's work on SNAP benefits and crime:

"SNAP Benefits and Crime: Evidence from Changing Disbursement Schedules" by Jillian B. Carr and Analisa Packham.

 

Transcript of this episode:

 

Jennifer [00:00:07] 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 Jillian Carr. Jill is an Assistant Professor of Economics at Purdue University. Jill, welcome to the show.

 

Jill [00:00:25] Thanks, Jen. I'm excited to be here today.

 

Jennifer [00:00:27] We're going to talk about your research on how staggering the distribution of public benefits like SNAP  that is, instead of giving them to all recipients all at once can reduce crime. But before we dive into that, could you tell us about your research expertise and how you became interested in this topic?

 

Jill [00:00:44] Sure, so I would say that my expertise is broadly in government transfer programs and crime, and sometimes I get to study those together, but sometimes I study them separately as well. And my interest in this intersection of these two areas began when I lived in Memphis as an undergraduate and people were blaming this increasing crime that we were experiencing around twenty seven on some housing policy.

 

Jill [00:01:07] And I remember thinking, you can get data. We can answer these questions, and I think I want to do that. So one of my very first papers I wrote was about Section 8 housing vouchers, and people just kept asking me, what do we know about SNAP and crime? Is there a similar paper that exists that you can give us the numbers from and tell us how your results compare? But when I looked at the literature, I just didn't really find that paper that I was looking for. And so my coauthor, Analisa Packham at the time was doing some work on food insecurity. And I said, you know, why don't we write this together? We'd written together before and so it seemed like a great sort of intersection between our interests.

 

Jennifer [00:01:43] Your paper is titled "SNAP Benefits and Crime: Evidence from Changing Disbursement Schedules," as you said, it's coauthored with Analisa Packham, and it is published at the Review of Economics and Statistics. So tell us about SNAP benefits. What is this program and who is eligible to receive this type of assistance?

 

Jill [00:02:02] Sure. So most people know about the SNAP program, but they might not know its name because most people just call it food stamps even to this day, despite the fact that we don't use that name for it anymore. They used to actually give people stamps, but now we distribute benefits on debit cards and so the old name of food stamps doesn't make as much sense.

 

Jill [00:02:23] But essentially the way that SNAP works is we give low-income families electronic benefits on a debit card once a month, and those benefits can only be used on certain goods at approved retailers. So the certain goods that they can use it on, they have to be food goods and they have to be food goods for consumption at home. So you can't use food stamps at a restaurant, for example. You can't use snap benefits at restaurants, for example. But it's actually not that specific about the kinds of food that you can buy, whereas some other programs are a little bit more specific about the types of food.

 

Jill [00:02:57] So in order to get SNAP benefits, your eligibility is mostly determined by family income related to the size of your family. So on the federal level, the benefits are only available to people who receive income around 130% or less of the poverty line. So here in Indiana, the gross income limit for a family we would like to receive SNAP would be $2,790 if they're a family of four and the max benefit here is around $646. But as families earn more, still than the limit, that benefit will be reduced. And on the average in the US, it's around $130 per recipient. This is not an insubstantial benefit to these families, but it is important to note that it has to be spent at grocery stores.

 

Jennifer [00:03:46] Okay, so SNAP benefits are typically distributed electronically. It sounds like this changed. So tell us more about the logistics of this program. When do people get the benefits during the month? It sounds like it goes on on a debit card, but just tell us more about kind of how this all works in practice.

 

Jill [00:04:06] Sure. So each recipient receives benefits only once per month. And this is because of a federal rule. But each different recipient can receive their benefits on a different day, so many states in the past just gave everybody their SNAP benefits on the first day of the month. But these days, most states spread it out over the month in some way, and many states have chosen ways of spreading out the benefits. That's somewhat random. So in many states, they assign a case number to every recipient. And what day of the month you get your benefits on might be determined by whether that last number in your case the last digit is even or odd, or whether it's a multiple of three or something like that.

 

Jill [00:04:51] And in this project, we'll be talking quite a bit about states where they distribute it by first letter of last name. They kind of decide based on that, all the A's get it on one day, all the B's get it on another day and such. So in some states, it's a spread out as the entire month and some keep - continue to do it on just a couple of days of a month. And those are typically states that have less recipients and low poverty levels.

 

Jennifer [00:05:13] And so why do states do that? So I think probably most people have in mind that, you know, people get a monthly check or a monthly deposit in their account or something. So why would a state want to stagger the benefits in this way?

 

Jill [00:05:29] So the fun, interesting back story that you find when you go digging in legislative briefs and stuff is actually that grocery store owners had a real problem with once per month benefits where everybody got them on the same day, that is. And the real problem was they had a hard time keeping their shelves stocked. They had a hard time with hiring decisions, things like that. And so grocery stores have consistently lobbied states, especially the ones with lots of SNAP recipients, to vary the timing of their benefit disbursement schedules more in order to alleviate that that pressure on grocery stores. So, you know, we're going to show that there are some benefits to staggering benefits in this way. But that's not actually the reason they do it's mostly for grocery stores.

 

Jennifer [00:06:17] And I also want to clarify what we're meaning - what we mean by staggering here. So here staggering means that different people will get their full benefit at different times. So you could also imagine a system where you get, say, a quarter of your benefit at four different days over the course of the month so that you're sort of spreading out an individual's benefits over time. That's not what places do. So in general, what we're talking about here is you get your benefits on the 5th, the month I get my benefits, then on the 10th, the month. Is that right?

 

Jill [00:06:46] Yes, exactly. And in fact, the former that you discussed, the idea of giving out part of everyone's benefits on multiple days - so you get half on the first and half on the 15th - that's actually not allowed on the federal level at this time. So if any state wanted to do that, they'd have to receive a waiver from the federal government to do so.

 

Jennifer [00:07:06] Huh? That's really interesting. Okay, so what are the mechanisms that we should have in mind as we talk about how this program might affect crime? What are the potential ways that SNAP benefits might affect criminal behavior?

 

Jill [00:07:19] Well, I think it's sort of the standard list that we think of when we're thinking about any kind of in kind of transfer where we're giving someone a good or a voucher for a good instead of just giving them cash, although a lot of the time they end up being basically the same kind of list anyway. But we could think about this as shocking a family's income resources and potentially changing their labor decisions. So when we think about it affecting their income - I try to think about it a little bit more broadly in terms of resources and not just food, because at the end of the day, a family's income is really fungible in a way, in the sense that if I were to give you a $100 gift card at the grocery store, you might buy more food at the grocery store. But unless you buy $100 more food, then you're likely displacing some part of your regular grocery store budget and then that income is freed up to spend on something else. So we tend to think of SNAP benefits as changing a family's overall resources, although to the extent that they are fungible, we're not totally sure how much they really see it as a cash transfer.

 

Jill [00:08:22] There's actually a bit of contentiousness in the literature about that, but we think about it as families have more resources overall. And how they use those additional resources really impacts how we think about their likely overall responses in terms of crime. So you can do good things with additional income of course, you can buy nutritious food, you can buy books, you can sign up for a night class that you have to pay tuition for all kinds of good things that even if not immediately down the road, could potentially impact criminal outcomes. But I think the the primary sort of upside to the income shock is that it might reduce someone's need to commit crime as well.

 

Jill [00:09:01] And this is something that we sort of find overall in our work. I like to call it the Aladdin hypothesis. So it's sort of the first scene in Aladdin or if you are a classier person than me, you can think about it as Jean Valjean, where someone's committing crime out of a really basic need to feed their family, for example. We also worry about the negative impacts of additional resources. These are something that is also very real in the literature where people with more resources can buy more alcohol, they can buy more drugs. And in fact, when families get benefits on the weekends, some literature has shown that they buy more beer at the grocery store while they're spending their benefits. Now, obviously, you can't use your SNAP benefits on beer, but if you free up $20 of your grocery store bill by using your SNAP benefits, then you could potentially use that $20 to buy something like beer.

 

Jill [00:09:51] And then on the labor side, it really just comes down to you are likely to decide to work less, and that's not, again, not necessarily a bad thing. And that's due to the structure of the program and the fact that you might need less income now that you have less earned income, that now that you have these additional resources. And so maybe somebody gets to quit their third job and they get to spend more time with their kids and that's great and we think that's wonderful. But they can also use additional leisure time to do things that are sort of bad, including crime itself, as well as, you know, excessive drinking and drug use and things like that. So so the end of the day, it's sort of the classic economists setup where we think it can really go either way. And there are a lot of potential mechanisms.

 

Jennifer [00:10:33] So as you and Analisa first started this project, what challenges did you have to overcome? So more broadly, what are the challenges in figuring out the causal effects of SNAP on crime?

 

Jill [00:10:44] Yeah, I think some of the primary identification challenges come from the fact that families who receive SNAP benefits are really different from families that don't receive SNAP benefits. And that's hard because if we compare people who are eligible to people who are not eligible based on income, they're clearly going to be very different. SNAP recipients will have lower education levels, they're more likely to be single parent families live in high crime communities. But if we compare the recipients to people who are eligible but not recipients for whatever reason, then the SNAP users are more likely to exhibit really positive attributes. They're likely to have stable addresses. They're not moving as much. They're more motivated to try to escape poverty and provide for their children. So they're not actually a great counterfactual either. So that's the primary challenge of thinking about some of these net effects of SNAP and data is really hard here as well, because who's receiving these programs is really sensitive. And we we need to get things linked up in order to consider SNAP with other outcomes such as criminal outcomes. And understandably, agencies can be really hesitant to do so. So for us, it was really about trying to figure out a way to find a source of variation that exists within the program that we can use to either identify individuals who receive versus one who don't or to identify behavior that signals that the recipients are actually doing something different than non-recipients, potentially.

 

Jennifer [00:12:12] So before this paper, what had we known about the distribution of SNAP and other public benefits on crime and criminal behavior?

 

Jill [00:12:20] So we knew that how states distribute their SNAP and TANF benefits overall can affect whether or when crime is happening. So if they spread it out, as we've been discussing and calling staggering versus giving out all of these benefits on one day can have sort of net effects of an overall city, crime patterns and this was a really important paper from Foley in 2011. And finding this paper encouraged us to think that the type of variation that we're using on benefit days and dispersement schedules could actually really have some some sizable effects because Foley was looking at sort of a static time in terms of policy and we were able to come in and see what happens when you see a change. And what we find due to the change is very consistent with Foley's work. But we also knew that that sort of cyclical patterns to crime and sort of negative behaviors was supported by other programs as well. So it's present for Supplemental Security Income and disability benefits as well.

 

[00:13:18] And there are a host of negative effects that are found in that literature, primarily that literature being driven by Evans and Moore and Dobkin and Puller. Those are the two primary papers there. But they're all finding that there are changes in things like mortality and drug related hospital benefits, all these sort of negative outcomes that we think are very similar in the mechanisms that we were considering. And then within the SNAP literature and sort of very close to the type of variation that we're using, there's some work by Elena Castellari and her coauthors where they show that recipients are buying more beer when they receive benefits on the weekend and I mentioned that earlier as well. And a similar sort of significantly overlapping group of coauthors, they also show that drinking and driving actually goes down when families receive their benefits. And so those who are, you know, a little they're playing sort of slightly different mechanisms there. But again, it sort of made us think that there is real scope for people to change their behavior and potentially in ways related to crime due to these benefit fluctuations.

 

Jennifer [00:14:22] Okay, so you're going to consider the effects of SNAP benefits in two states, Indiana and Illinois. Where are the details of how benefits were distributed as well as recent changes to those policies both provide nice natural experiments that allow you to isolate the effects of snap distribution on crime. So let's talk about Illinois first. What was the distribution policy there and how did it change?

 

Jill [00:14:47] So Illinois had frequent changes to their snap distribution schedules in the last 20 years, but that the time that we're looking at, they had sort of on the books a pretty large set of distribution days, but basically all regular SNAP cases that weren't through special side programs were receiving their benefits on the first day of the month. So what they change was they started giving out more benefits on the fourth, the 7th and the 10th out of that sort of general snap population. And this sort of full set of dates on the books goes through the 23rd. But again, most of the new distribution is going on, on the 4th, 7th and 10th. So we think about it as sort of a like just a little shock to how they're spreading out the benefits over that first sort of third of the month.

 

Jennifer [00:15:31] Okay, so then how are you using that policy change to measure effects on crime?

 

Jill [00:15:35] So we're using an empirical strategy that's going to leverage the really discrete timing of the policy change. So in that way, we'll be doing a regression discontinuity type model, it's a little bit like an interrupted time series as well, you can sort of choose which way you want to describe it, but empirically, you sort of average them the same way. And that we look at the time just before the schedule change and we compare it to right after the schedule change. And we're going to look for a discrete change in both the amount of crime, the locations, the timing, all these kinds of things. And we control for a whole lot of time specific effects here. We're thinking about things like day of week, for example. We know that crime is always higher on weekends. Right and things like that. So we're controlling for all of these kinds of things. And that's sort of our simple approach.

 

Jennifer [00:16:23] Okay, and so just to make it even more intuitive, you've got a month just before the policy change where most people were getting their benefits on the first. And then you've got a month just after the policy change where the benefits are staggered throughout the month. And if we think that staggering matters, then we think that crime should either fall or increase right after the policy change. And that's basically what you're going to see. You can kind of imagine graphing over time the trends in crime you expect either a drop or an increase right after that date of the policy change. Does that sound like getting that right?

 

Jill [00:16:56] Yes, exactly. So if the policy had no impact and we were to just graph crime over time, it would just look flat or it would be following some type of trend, a seasonal trend or something. But we're what we're really looking for here is some type of really discrete change at that same time. And going into this project, we weren't sure if that discrete change would occur in the overall levels of crime or if it would really be concentrated in the days where these changes are happening. So as you take people away from the first, we might expect crime to go down on the first. If you think that the days later in the month we're going to see potentially less crime as people have less need than we would expect decreases there as well.

 

Jill [00:17:37] We weren't really sure if this was going to be a shifting story or a change in levels. And what we ended up finding was really a change in levels where, you know, for any day of the month, it's essentially falling.

 

Jennifer [00:17:47] Okay, so what data are you using from Illinois to dig into all of this?

 

Jill [00:17:53] So we have three different data sets - I guess, four maybe - that we're working with from Illinois and the first is a really great data set that's actually available online. The city of Chicago puts their crime reports on their city data website, and they're incredibly detailed. And so this is an incident level data set. So any time that a police officer makes a report - takes a report - it ends up on this website and each observation as the incident itself with a location, the location is really critical for what we are doing and that it has either, well, it has for every single one of these incidents the latitude and longitude or if you prefer, it has a block level address. So both of those are available. We prefer the latitude-longitude because we want to get up to Census tracts. It's a lot easier that way. But the officers also record a description where they say this incident occurred at a residence or at a dormitory or at a grocery store or in our case or a park. And so we're able to use those as well to say we know that this event occurred at a grocery store.

 

Jill [00:19:01] So also about each individual incident, we know when it occurred, we actually don't even need the super specific time details here. But it does have them. We'll just look at the day and then we also know what type of offense, what type of offense was recorded. So we'll know if it was a theft or a battery or what have you. And that ends up being really important to us. The other data set that we use that's a little bit more administrative in nature is we were able to get the state of Illinois to give us data on the redemption of benefits. And so every time somebody uses SNAP benefits at the grocery store, that goes into a tally. That gets reported back to the state, and so at the end of the month, the state has all these reports that say, okay, this many dollars was redeemed on this day at this grocery store. And so, I mean, obviously has a role in the reimbursement of the stores for the benefits. But it's also really cool because we're able to look at this on the day level and say, okay, well, when this change happened, how many people would have normally spent benefits on the first day of the month versus the second versus a third? And we can see sort of those those patterns. So that's the second data set we used.

 

Jill [00:20:12] And then we also had a list of official SNAP retailers. So we know which places are licensed to take SNAP benefits. And then we also brought in some census data from the American Community Survey to look at demographics for different Census tracts because we wanted to think about where what locations are likely to have the greatest effects.

 

Jennifer [00:20:33] I want to go back to that, the reimbursement data really quickly and just to kind of emphasize that. So you think about kind of standard economic theory, about how people would use their benefits, the most basic models would predict that there would be no change in when they actually use them. Right. Because everyone is a perfectly rational consumption, smoother and they know that, you know, they should they only get this much money for the month and they should, you know, carefully calculate exactly how much to spend on every day. So they still have money left at the end of the month. And of course, we generally think that that model doesn't quite fit reality, but you'll be able to test that.

 

Jill [00:21:12] Yes. So that that's a really crucial component to really finding anything at all when you're looking at these disbursement schedules and when you're thinking about staggering benefits, is that if people really do consumption smooth perfectly the way that as economists we suggest they should, then you really should never find anything in any of these papers. And so being able to sort of show that the dispersement does or dispersement dates do actually impact when people spend their benefits. That's that's a pretty crucial component to any of these SNAP papers.

 

Jennifer [00:21:47] Right. Okay, so let's talk about the results. What do you find are the effects of staggering SNAP benefits on both when those benefits are used and then on crime rates?

 

Jill [00:21:57] Yeah. So, again, thinking about this redemption of benefits, we find that the benefits end up getting use later in the month when there is a staggered system in place, which is what we'd expect. And it's a much smoother pattern without this big spike right at the beginning of the month. So before the change, around 18% of benefits were redeemed on the first two days of the month. And then after the change, it fell to only 9%. So we're seeing a pretty big change in terms of spreading out the use of benefits over time. And then we think about the fun stuff, right? The crime and theft that we're interested in here. When we focus on these events occurring at grocery stores, crime falls by around seventeen 17.5% and theft falls by around 20.9% after the policy change at grocery stores. And that's across all days of the month.

 

Jennifer [00:22:46] That's a huge change in overall theft at grocery store rates just by spreading out when people get their full benefits.

 

Jill [00:22:56] Yes.

 

Jennifer [00:22:57] And you also see changes in - you had originally hypothesized that the the crime might just move around. Is there some of that or does that really just kind of fall on all dates?

 

Jill [00:23:09] Yeah, so there definitely is some of that as well. And and what we find is that there is a little bit less crime at grocery stores on the first day of the month, and then there is remarkably less around days like 14 to 23 relative to when SNAP was in the first. And the when we think about theft at grocery stores than the first, the month doesn't really have any change. But that change between maybe the 14th and the 23rd where there's a lot lower crime after the policy change is echoed there as well. So basically what we're seeing is less crime and theft committed at grocery stores between the 14th and the 23rd of the month, which is when we believe that most families are really running out of benefits. So it's this really crucial time for families as they run out of their SNAP benefits. And that's where we see the big change once they start distributing them later.

 

Jennifer [00:24:05] You also mentioned the potential that people could be spending some of this extra income now on things like alcohol that could increase crime. Are you able to see that in the Illinois data at all?

 

Jill [00:24:18] So in the Illinois data, we don't see that a whole lot here in this work. We did look into this a little bit more for a follow on paper where we're looking at domestic violence. And there we do find that drug use or crimes related to drug use appear to be going up as well. So there could be a little bit of these negative effects, but they're definitely not nearly on the magnitude of the — result.

 

Jennifer [00:24:43] Got it. Okay, and then you also are able to consider how the effects vary across different neighborhoods because you have the latitude and longitude of where all these thefts occurred in this specific grocery stores and everything else. So what did you find there?

 

Jill [00:24:58] Well, we really thought that it made sense to do this analysis, because if you're finding larger effects in places that nobody really is enrolled in the SNAP program, that would be a bit weird and a bit surprising. So so, yeah, we split things into places that were high and low SNAP enrollment and we found that generally across the board there are larger effects in places with high SNAP enrollment. And in fact, when we look at crime at grocery stores and theft or grocery stores, which are you'd say the primary results, the paper, and we actually find that there's no discernable effect in a statistical sense on the low SNAP enrollment places, but there is in the high SNAP enrollment places. So the effects seem to be driven by the high SNAP enrollment. In fact, instead of being 17.5 and 20.9% effects, we're finding 24% of a decrease in grocery store crime and 32% decrease in theft. So the results are actually quite a bit larger when we sort of split them in that way and then we also wanted to look at Census tracks with SNAP retailers, which is almost mechanical in a way, because all grocery stores are generally approved, SNAP retailers, in fact, I would say that many CVS and various types of drugstores are also on that list, as well as things like 7-Eleven, gas stations, any type of little corner store. So we really think that those are the kinds of places where we could see effects and we do, in fact, see more effects in places that have more SNAP retailers. They're much larger. In fact, the the effects on low SNAP retailer locations are very close to zero.

 

Jennifer [00:26:37] Great. Okay, so let's talk about Indiana now. So how were SNAP benefits distributed in that state and what changed there?

 

Jill [00:26:45] So Indiana is a really neat place. I'm not just saying that because I live there. I'm currently an Indiana resident, but the way that they give out their benefits is really interesting. And they've sort of been on this boat about staggering benefits for quite a while. So we study a policy change that happened in 2014, but Indiana had staggered benefits before the policy change already, they just made them more staggered.

 

Jill [00:27:10] So the older system, they staggered benefits between the 1st and the 10th day of every month by first letter of last name. So, for example, under the old system in Indiana, I would get my benefits on the second day of the month because my last name begins with the letter C, and actually you'd get yours on the second as well, Jen, because you're A, D, C and D got theirs together in Indiana. So I'll use my coauthor Analisa an example here, because her last name begins to the P, so she would have gotten her benefits on the seventh day of the month under the old system. So that's sort of the gist of how we're going to think about each individual sort of benefit month schedule. It's like we start the month for each person on the day that they get their benefits. But it makes Indiana even cooler is that they started out in 2014 where it's sort of a larger swath of the month. So they kept the same number of dispersement days, but now they give them out on the odd days between the 5th and the 23rd of each month. So after 2014, you and I would get our benefits on the 7th, but Analisa her date would be pushed all the way back to the 17th. So because it's so spread out, this benefit month, in fact, relative to the calendar month is actually really different. It's really stark, which is which is really useful for the way that we will create our models.

 

Jennifer [00:28:29] And so what does this policy change and the dispersement schedule in in Indiana, what does it allow you to test that you couldn't test in Illinois?

 

Jill [00:28:36] Right. So we're actually not focusing as much on the policy change, in the Indiana analysis, but we're more so using it in order to more precisely identify the effects that we're most interested in. So we could look at the timing of these offenses that will we'll have in our data set for individuals relative to their distribution date and so instead of looking at it sort of on the much sort of higher level where we're looking at aggregates, we're looking at individual outcomes now, and because we know the first letter of individuals last names, then we can map them to their distribution dates. So it allows us to ask questions like do people commit more offenses at the end of the benefit month when we think they're out of resources? Or we could also consider, do they commit more at the start of the month because they feel that they are resource rich?

 

Jennifer [00:29:23] Right. And so you be looking at, for instance, do I? So my last name starts with D, so and under this new system, I get my benefits on the 7th, if I remember correctly and so am I more likely to commit crime on the 6th, say, before I get my benefits, whereas Analisa is getting her benefits on the 17th, we might expect her to commit crime on the 16th.

 

Jill [00:29:44] Exactly right. And that's the thought is that if we find those kinds of effects consistently across letters and across distribution days, then it would point towards that being the mechanism and not something else. Like people tend to get paid on the 15th. Well, people get paid on the 15th and that affects Analisa on the 16th. We wouldn't expect to see the same thing for you on the 6th if that was the only effect.

 

Jennifer [00:30:05] Right. And so the underlying assumption here is that the the date of when we're getting benefits is not correlated for other reasons, like payday's with when when we might otherwise want to commit crime. So if people also got. Yeah. Their paychecks on the schedule, then that would be an important confounding variable. But we don't think that's the case. And in general, it seems reasonable. That seems pretty random in terms of kind of last names and and when when these benefits would take place. You do talk about in the paper that you thought about whether different races or ethnicities might have names that start with similar with the same letters. You looked into that a little bit. Do you want to talk about about what you find there?

 

Jill [00:30:48] Sure, yeah. I mean, yeah, we definitely do not believe that first letter of last name is unrelated to other personal characteristics. We can see that in this data set - we've worked on other data sets where we're using a similar identification strategy - and it is 100% not true there either that first letter of last name is random, but what we really need to believe is that it is not related to the timing of the benefit disbursement. And what's really great about Indiana and this policy change in Indiana is that for each letter we have two different disbursement dates, which means that we can give them a fixed effect as well for being a C/D in Indiana, right, whereas Analisa would have a fixed effect for being in a I think it's just a handful of letters around P I'm trying remember, I think it's four of them so she would be in sort of her letter group fixed effect.

 

Jill [00:31:44] We would have our own letter or group fixed effect. And in fact, in Indiana we actually know the exact letters. So she could just get a P fixed effect. You can get a D, I can get a C and we can all have our own letter fixed effect essentially, and that'll take care of anything. We think that is different within those letters and we also can control directly for raciness as well. So we're controlling for race. We're controlling for sex,  and we're controlling for these letters fixed effects and so given that each letter gets two different disbursement dates in our data set because of the policy change they were, then we're confident that we're going to be able to control for any of these, you know, letter specific problems, especially if you're worried about certain letters, always having dates that are related to, you know, paydays or TANF benefit dispersement or something like that. Since we move them to another date, we can kind of get away from that.

 

Jennifer [00:32:31] Yeah, it really is a very nice natural experiment. Yeah. Okay, so what data do you have from Indiana to look into all of this.

 

Jill [00:32:40] So are data from Indiana came to us from the Indiana Department of Correction. So we worked with them. We applied to get this data set from them and they gave us a data set of convictions that resulted in the Department of Correction being involved in some way. So every conviction that results in probation or incarceration ended up in our data set. And so what we know from the Department of Correction data is the first letter of the individual's last name and then we know the date that they committed the crime, which is really critical because the date they're convicted might have nothing at all to do with their SNAP schedule. So we really needed the date the crime was committed. We also needed to know the type of crime and then they provided us with the race and the sex of the individuals as well as the county of conviction. So we had a handful of details about the convictions themselves.

 

Jennifer [00:33:29] Okay, so let's talk about the results here. What do you find is the effect of receiving SNAP on criminal behavior in this context?

 

Jill [00:33:36] So we find overall that individuals are more likely to commit crime at the very beginning and the very end of the benefit month. We think this is coming from sort of those two motivations that we talked about earlier, where at the beginning of the month, people are resource rich. In a way, they're going to be more likely to be out and about, which is a huge part, actually, not just about having resources, it's about interacting with other humans. It's inherently dangerous turns out you're more likely to be you know, your mortality goes up, you're more likely to be assaulted, all these kinds of things. But at the beginning of the month, your behavior changes in ways that make you more likely to commit crime.

 

Jill [00:34:11] Then at the end of the month, the same thing happens, but for very different reasons. And I think to me, the most striking result from this part of the paper is that these effects that the end of the month are actually driven by theft committed by women and so there's not a whole lot in the economics of crime literature where we talk about women. And I thought that was just really interesting and also very fitting because we know that many SNAP recipients are single moms and so it makes sense in terms of which group is likely to be impacted.

 

Jill [00:34:40] But it also is a little bit disappointing - dismal, I guess - in the sense that if these women are being convicted for theft, they committed in order to feed their kids, then they're potentially losing these benefits because there are rules about criminal history, but there's also a chance that they're then becoming incarcerated. And if they're now incarcerated, those kids are potentially in foster care or having to live with grandparents and things like that. So you're taking potentially parents out of the home as well, which is a little bit of an even greater problem.

 

Jennifer [00:35:13] Right. Okay, so we've got results from two different states now. So putting it all together and just kind of summarizing everything you just said, what's the punch line of these two analyzes? What are the main takeaways?

 

Jill [00:35:24] So I think, you know, one simple very punch line take away is that staggering benefits appears to be generally a good idea, although we are continuing to put together some of the information across our own work and then other other authors work to sort of get the full picture in terms of the costs and benefits of moving from a one day system to a staggered system.

 

Jill [00:35:46] But I think that most of what we learn kind of comes down to the fact that people are just running out of benefits and whether that's this issue with consumption smoothing or whether it's an issue just that the benefits aren't big enough, we can't really tell empirically, but one of those two things is likely causing this these effects that we're finding.

 

Jennifer [00:36:04] And has any other research come out since you first released this paper, that adds to our understanding of how public assistance can affect criminal behavior?

 

Jill [00:36:12] So I'd say there are two studies that were sort of contemporaries of this one that get it sort of different aspects, but I'll think about SNAP crimes. So Andrew Barr, who is one of your colleagues with his coauthor, Alex Smith, did some work on childhood access to SNAP and its effect on criminal convictions later in life as adults. And so they find that childhood access greatly reduces criminal conviction as an adult, which is really exciting work. And then Cody Tuttle has done some work thinking about SNAP for felons, and he finds that the ban on letting felons receive SNAP increases recidivism among drug traffickers. So it seems like when you take away the ability to get SNAP benefits from potentially high risk populations, there are more likely to commit more crimes. And we can think about sort of the way that that fits into the decision making process for a potential criminal in a very Beccarian way if we wanted to. But I think it says that both of these papers together and and ours as well, are showing that SNAP can really reduce the amount of crime that that's going on and that's out there.

 

Jill [00:37:20] And then I'd also say that the literature on SNAP schedules and whether to start to stagger benefits has really blown up around the time that we started working on this. There's a pretty big health literature these days thinking about how people utilize medical resources related to SNAP schedules. And then there's also an exciting and sort of burgeoning education literature as well, where researchers, including Analisa and I and another project are thinking about how high stakes testing is impacted by snap benefits schedules because some kids are showing up to take a test on a day where they had a great dinner the night before and their families had resources for the last week or so and then other kids are showing up during this time of sort of extreme resource scarcity and taking these high stakes tests. And so so all these literatures have sort of really gotten a lot of attention and gotten very exciting recently. So it's it's a good it's a good role to be in.

 

Jennifer [00:38:17] So what are the policy implications of your results in this paper and other work in this area? I'm sure people come to you now and as or they certainly will if they haven't yet asked you what to do about SNAP, what do you tell them?

 

Jill [00:38:30] Well, I think normally when I get the chance to talk to policymakers about this, I kind of go back to the two possible driving mechanisms behind the extreme scarcity that families experience. So, again, it's this combination of people being bad at budgeting, which it's not just SNAP recipients, it's everyone's bad at budgeting. And the fact that if you don't give people enough benefits to begin with, even if they do budget really well, they're still going to experience extreme scarcity. So typically, when I think about policy implications, I think about how we can address either or both of those potential causes of the scarcity. So I tend to tell policymakers, let's test them. I think that the easier one to test might be giving out benefits twice a month to families, you know, splitting their benefit into two disbursements, but the other thing that would be really cool to do would just be to increase how much we give people. So those are the two two potential policy implications, I think is let's test those two different mechanisms.

 

Jennifer [00:39:28] And so that's part of the research frontier for sure. Are there other big questions in this area that you and others need to be thinking about going forward?

 

Jill [00:39:37] Well, I do think that all these sort of other outcomes that are impacted by SNAP are really important in the sense that crime is connected to so many of these things. But I think that for me, the real research frontier is probably really thinking about access to the benefits. There are a whole lot of challenges to studying SNAP, and I think that identification of causal effects is even harder. When we think about giving access to SNAP in our modern setting, we look back at the original SNAP literature on the roll out of food stamps that is a great way to study a program like this and the impacts that it has.

 

Jill [00:40:18] But we don't have a great equivalency in our modern SNAP era. So I think that looking at sort of this, you know, extensive margin as far as who gets benefits and how they qualify for benefits is maybe maybe the really important policy question. Although obviously the how much do we give people? When do we give it to them? All very important. But I do think the you know, what happens if we give more people benefits is is the really, really essential question that I'd like to see answered.

 

Jennifer [00:40:47] My guest today has been Jillian Carr from Purdue University. Jill, thanks so much for doing this.

 

Jill [00:40:52] Thank you for having me, Jen.

 

Jennifer [00:40:59] 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 and thanks also to our Patreon subscribers. This show is listener supported, so if you enjoy the podcast, then please consider contributing via Patreon. You can find a link on our website. Our sound engineer is Caroline Hockenbury with production assistance from Elizabeth Pancotti. 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.

Jennifer DoleacRegression-Discontinuity, Resources