Episode 48: Cody Tuttle

 
Ep48_Tuttle_photo.jpg

Cody Tuttle

Cody Tuttle is a postdoctoral fellow at Princeton University.

Date: March 30, 2021

Bonus segment on Dr. Tuttle’s career path and life as a researcher.

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Dr. Tuttle's work on how eligibility for SNAP benefits affects recidivism:

"Snapping Back: Food Stamp Bans and Criminal Recidivism" by Cody Tuttle.

OTHER RESEARCH WE DISCUSS IN THIS EPISODE:

 

Transcript of this Episode:

Jennifer [00:00:08] 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 Cody Tuttle. Cody is a Post-Doctoral Fellow at Princeton University, and he's joining the Economics Department at UT Austin as an Assistant Professor in the fall. Cody, welcome to the show. 

 

Cody [00:00:29] Thanks for having me. 

 

Jennifer [00:00:31] Today, we're going to talk about your research on how access to public benefits like food stamps affects recidivism. But before we get into that, could you tell us about your research expertize and how you became interested in this topic? 

 

Cody [00:00:42] Sure. In terms of my research interests, I do applied micro work in public economics, and my research has mainly focused on trying to understand the roots of economic and social inequalities across groups and to provide evidence on how government policy can either reduce or exacerbate those inequalities. Recently, I've been particularly focused on government responses to crime and racial inequality. That's the broad motivation for why I study these topics. 

 

Cody [00:01:06] As for this paper in particular, I really first became interested in SNAP in general. Just for background, SNAP is the Supplemental Nutrition Assistance Program, commonly known as food stamps. And it's a means tested in kind transfer program that helps people with low income pay for food. It's a program the government spends over $70 billion on in a given year, and it covers millions of people. But in reading more about it, I learned that despite aiming to help people in vulnerable situations, one vulnerable population that's banned from receiving SNAP benefits at the federal level is people convicted of drug felonies. Now, a lot of states have repealed all or part of that federal ban, but 26 states still have some version of the ban in place. This is especially concerning because reentry and recidivism are big issues in the US. Over two million people are currently incarcerated in this country, and we have one of the highest incarceration rates in the world. Of course, when you lock that many people up, you eventually release many of them. And in 2019, over 600,000 people were released from prison and charged with this very difficult task of successfully reintegrating into their community. 

 

Cody [00:02:08] Of course, you have to wonder how these people turn out. The answer is not good. By some estimates, over three quarters are rearrested within five years and over half wind up back in prison. So we have this really big problem with prisoner reentry in the US and a large government program that's supposed to help people who are facing hard times. So I think it's a natural question to ask whether providing people with access to food stamps or financial assistance more broadly will reduce recidivism. And I think it's an important policy question to ask because again, so many states still have some version of this ban in place. 

 

Jennifer [00:02:39] So your paper is titled "Snapping Back: Food Stamp Bans and Criminal Recidivism." It was published in 2019 in the American Economic Journal: Economic Policy. And in this paper, you focus on an element of welfare reform in the 1990s. So tell us about that reform and the piece that affected food stamps specifically. 

 

Cody [00:02:57] That's right. The broad reform was called the Personal Responsibility and Work Opportunity Act. Or if you sound out the initials, we often call it PRAWOA. And it was passed in 1996. It made some pretty dramatic changes to welfare policy in America. A lot of which were motivated by this idea that people who receive government transfers should be encouraged to join the labor force and reduce the reliance on transfer programs. One relatively small provision of that bill, or small in relation to the other sweeping changes it made, was that it imposed a lifetime ban from food stamps on people convicted of drug felonies. One additional detail here is that it also banned people convicted of drug felonies from receiving Temporary Assistance for Needy Families, or TANF, which is a cash welfare program. That said less than 10 percent of non-elderly adult TANF recipients are male. And in the data that I use, almost 90 percent of people convicted of drug trafficking are male. On the other hand, in some years, almost 40 percent of non-elderly adult food stamps recipients are male. So I always say the paper is studying the effect of banning people from food stamps. But you could think of the effects here as some combined effect of the TANF and food stamps ban. 

 

Jennifer [00:04:03] And what was the motivation behind this ban on food stamp access to people with certain criminal records? What was the goal of that policy change? 

 

Cody [00:04:10] That's a good question, and I'm not really sure I know the answer. I know that the ban was added to the bill by Senator Phil Gramm on July 23rd, 1996, a month before the bill was signed. And if you look at the congressional record at the time, he basically says if we're serious about our drug laws, we shouldn't give welfare benefits to people who are violating our nation's drug laws. So in some ways, it seems like it was born out of this war on drugs mentality that people committing drug crime were doing something so bad that they didn't deserve government assistance. And in the congressional record, there's no discussion about this ban having a deterrent effect, so it doesn't seem like it was motivated as a crime control mechanism. 

 

Cody [00:04:48] Likewise, there's no discussion of its effects on recidivism. As far as I can tell, it was mainly punitive. If you read some defenses of the ban now, it seems some people believe that the ban prevents benefits from flowing to people who might be making some illicit and unreported income. But that's basically assuming that when someone leaves prison, they'll automatically go right back to selling drugs, but the idea of this paper is that by providing financial support, we might be able to prevent that. 

 

Jennifer [00:05:13] So in this paper, you focus on Florida, which amended this ban slightly. What policy went into effect there? 

 

Cody [00:05:19] Yeah. One key element of this federal ban is that states were allowed to modify the ban or opt out completely. As an aside, that was really a theme of PRAWOA more broadly, the importance of state control over their welfare systems. And it was applied, in this case too. Florida was a state that modified the ban, and they changed it to apply only to people convicted of drug trafficking. That means they allowed people convicted of possession or lower level selling offenses to receive food stamps upon release. The final detail here is that the ban in general only applied to people who committed their drug trafficking offenses on or after August 23rd, 1996, which is the date PRAWOA was signed. 

 

Jennifer [00:05:56] And you have a footnote in the paper about why you picked Florida, which I think will be of interest to listeners who wonder how the research process works. So it sounds like you also got data from North Carolina, but weren't able to include it in your analysis. Why not? 

 

Cody [00:06:09] Well, first there were two key requirements for selecting a state to study. The first was that the data include the day the offense was committed in the eyes of the justice system. A lot of publicly available criminal justice data include the date of conviction, but conviction typically occurs weeks or months after the person has committed the offense and is arrested. The second requirement is that the state legislature had maintained some portion of the SNAP ban. At the time, I was working on this project, the only public data I was aware of that met these requirements was from Florida and North Carolina. Both of these states actually implemented the ban in similar ways. Essentially banning people convicted of drug trafficking offenses, but not banning people convicted of lower level selling or possession offenses. 

 

Cody [00:06:49] To explain why North Carolina wasn't a viable option, let me step back just for a second and give some detail about the research design. The critical assumption of the research design in this paper is that people who commit their offenses just before August 23rd, 1996 are similar to people who commit the offense just after August 23rd, 1996, except that those people who committed just after will be banned from SNAP upon release and those who commit it before will have access to SNAP. Of course, this isn't really a trivial assumption. You know there are a number of things that could go wrong here, so to speak. For one, people who commit offenses could become aware of this ban and reduce their level of criminal activity. It's also possible that police, prosecutors, defense attorneys or judges become aware of the ban and change who they arrest, how they charge cases, how they make plea deals or how they sentence. In either of these states, for example, it would only take one large prosecutor's office to decide to change their charging policies in order for the composition of offenders to really differ before and after the policy. Finally, another wrinkle here is that the state could be changing other criminal justice policies around the same time, which would make it hard to identify the effect of the ban separately from the effect of these other policy changes. 

 

Cody [00:07:57] As researchers, we can't always observe all of the potential things that could be changing, so we rely on a variety of tests that indicate whether something else might be going on. For example, if a prosecutor's office were to start charging these offenses less, then we should see a decrease in the overall number of drug trafficking offenses after August 23rd. The data in North Carolina unfortunately, don't pass some of these basic tests. For example, there are fewer trafficking offenses in North Carolina after August 23rd, 1996 than just before. The tougher question is why exactly this happens in North Carolina? Again, it could be people who commit these offenses changing their behavior. It could be prosecutors or defense of counsel in North Carolina changing their behavior. Or it could be another coincidence state policy change. In fact, in June 1996, the North Carolina Legislature passed a bill that changed how class age and I felonies are handled in district courts, and drug trafficking in North Carolina is a class age or class I felony. So for a number of reasons, it seemed that North Carolina was not an appropriate setting to employ the research design from this paper. 

 

Cody [00:08:57] Florida, on the other hand, was an ideal setting. For one, and I'm sure we'll go into more detail later, but it did pass these standard tests that suggests nothing else was changing and that the people who committed trafficking right before August 23rd were very similar to people committing it just after. In addition, when I searched for news articles about this policy in Florida, I don't find any mention of this ban in Florida newspapers in the year before or after August 23rd. Likewise, I don't find any other major criminal justice reform passed at the state level in the months just before or after August 1996. Finally, Florida is also a great place to study this policy because aside from Texas and California, it has both the largest population in terms of people incarcerated and the largest population in terms of people receiving SNAP benefits. So this is a state where these issues are very present. 

 

Jennifer [00:09:42] Okay, so before this paper, what did we know about the effects of higher incomes or public benefits on recidivism? 

 

Cody [00:09:49] Well, there are definitely too many papers to talk about in detail. I'm bound to skip something important, but I would say first, there is this suite of three or four papers in the 1980s that mainly use randomized controlled trials to study the effect of giving people unemployment benefits when they're released from prison, and they find somewhat mixed results on overall re-arrest. But generally, they find suggestive evidence of decreases in re-arrest for property crimes or theft. I think those papers left a lot of people wondering if financial assistance was actually an effective way to reduce overall recidivism. In recent years, the literature is mostly focused on either the effect of labor market conditions on recidivism or the effect of other social programs. For example, Crystal Yang has a paper which suggests being released into a local labor market with higher wages, decreases recidivism and another work, Kevin Schnepel finds that people released into labor markets with a high availability of good jobs, things like manufacturing or construction work, are also less likely to end up back in prison. 

 

Cody [00:10:47] In terms of social programs, Emily Owens and Mike Lovenheim find a ban on receiving financial aid for college increases recidivism among people convicted of drug offenses. And in a 2015 paper, Ignacio Munyo and Martin Rossi find that when a prison in Uruguay increased the amount of money it gave people on the day of their release, what's commonly called gate money, it reduced the likelihood that someone would get out of prison but then commit a new offense that very same day. Finally, around 2017 to 2018, about the same time that I was working on this paper, a burst of other work came out that focused on the relationship between food stamps and crime or recidivism. Jillian Carr and Analisa Packham, who have both been guests on your podcast, for example, show that crime drops right after SNAP benefits are dispersed. And in a highly related work, Crystal Yang also studies the effect of the SNAP ban, but takes a different approach than I do. She uses the fact that states have opted out and modified the SNAP ban over time to study the effects of the ban on one year of recidivism rates. Essentially, she compares recidivism in states with and without the ban, both before and after those treated states remove the ban. And then on top of that, she also differences out any changes in recidivism in those states for offense types not subject to the ban. She finds it being released in a state that has completely removed the ban reduces recidivism in the first year of release about 2.2 percentage points, which is a 13 percent decrease from the overall mean. 

 

Jennifer [00:12:08] So how should we be thinking about food stamps, now known as the SNAP program, as you mentioned, as an income source? How much are monthly SNAP benefits typically, and what do we know about how common it is for people with criminal records to receive them? 

 

Cody [00:12:22] Yeah, I think there are two things worth emphasizing here. So first, for people who receive SNAP, those benefits are an important part of their overall income. Looking at a sample of men aged 18 to 65 who receive SNAP in Florida, in 2010 dollars, the average benefit they receive is about $150 to $200 a month. For those who report an income, the benefit represents about 20 percent of their reported gross income. So this is a substantial chunk of their overall resources. Second, based on survey evidence, it does seem as though SNAP is an important resource for people who are coming out of prison. The Boston Reentry Study, for example, surveys former inmates and finds that 70 percent report receiving SNAP benefits even just two months after release. 

 

Jennifer [00:13:03] And so what are the mechanisms you have in mind for why access to SNAP benefits might affect recidivism? 

 

Cody [00:13:08] I think there are a few different ways to think about this problem, but one way is to consider a basic labor supply model. And from that model, we get the prediction that taking away government transfers will incentivize or encourage work. The problem is that people who have a felony conviction face a ton of different hurdles to finding legal work. Employers are generally less interested in hiring them because of their conviction alone. They're subject to lots of different occupational licensing restrictions. They've been out of the labor force for years, etc. And not only do they face obstacles to finding legal work, they're particularly well positioned to find illegal work. They were doing illegal work before prison, and they may have made connections in prison that strengthen their attachment to the illegal labor market. So in this sense, the food stamps ban is a very strange policy because taking away those transfer benefits may encourage people to go back to the illegal labor market to make up for those lost resources. 

 

Cody [00:13:59] At the same time, there's probably even more going on here. For one, even just knowing that they have access to this safety net might change the type of decisions that a person makes, even if they never end up using it, they know it's there in case they experience a negative shock. But second, if someone who's banned from food stamps joins a household that is eligible, the household does remain eligible, but they don't receive any additional benefits. This means the overall value of their benefits is diluted because their household sizes increase, but their benefits have stayed the same. Things like that can put a strain on family connections that might be especially important for people coming out of prison. So I think this labor supply mechanism is important here, but I don't want to say it's the only story. 

 

Jennifer [00:14:36] I guess you could also imagine stories about if you're hungry, it might be more difficult to make smart decisions about what you're doing in a certain situation or something like that. So the actual nutritional benefit could be important here, too, I think. 

 

Cody [00:14:50] Yeah, I think that makes sense. That could be something about cognitive load and getting a good meal then maybe you're less likely to make rash decisions and commit a crime. 

 

Jennifer [00:15:00] So what are the hurdles that researchers like yourself have to overcome in order to measure the causal effects of SNAP access on criminal recidivism? Are there data challenges here or identification challenges or is it both of those things? 

 

Cody [00:15:12] I think there are a few different challenges. The first is a data challenge. It's not trivial to link administrative criminal justice data with administrative records on SNAP receipt or administrative records on income. There are some very cool projects that have done this or are doing this, but at the time I was working on this paper, it did not seem like a viable option for me, especially for a period of time going back to 1996. This means that if you have data on people released from prison and their recidivism, it's not easy to say who's eligible for SNAP or who's received SNAP. The policy that I study helps in that respect because the date of the offense clearly defines a group who's not eligible. Unfortunately, I still don't see a measure of SNAP receipt. So I end up estimating the effect of access to SNAP rather than the effect of receiving SNAP benefits. 

 

Cody [00:15:57] The second challenge is identifying the causal effect of access to SNAP. Even with ideal data that indicated who was eligible and who received benefits, you can't just compare people who are eligible to those who aren't eligible because those people who are eligible will be different in other ways. They likely have lower income, for example, and those income differences may affect recidivism. Likewise, you can't just look at people who are eligible and compare those who are receiving SNAP to those who aren't because they'll be different in other ways, too. For example, one possibility is that people are eligible for SNAP, but not receiving it, may be earning some illegal or informal income, like getting paid off the books that reduces their willingness to go through the application process. And so the policy helps in that respect because it imposes this access restriction in a fairly arbitrary way that gives us two comparable groups to study. 

 

Jennifer [00:16:43] So you're going to use the details of which offenders this policy applied to as a natural experiment. So walk us through your approach in a bit more detail. What's the ideal experiment that a researcher would like to run to measure the causal effect of this SNAP ban? And how does your approach approximate that experiment? 

 

Cody [00:16:59] I think the ideal experiment here is that we have two identical groups of offenders and we allow one group to have access to SNAP, but not the other group. Then we compare recidivism rates for these two groups, and the difference is the average effect of access to SNAP on recidivism. What happened in reality is that the policy in Florida applied the SNAP ban only to those people who committed a drug trafficking offense on or after August 23rd, 1996. So the intuition and again, the key underlying assumption, is that people who happen to commit an offense shortly before August 23rd are similar in all other ways to people who commit an offense shortly after August 23rd, except those people who commit it just after are going to be banned from SNAP once they're released, and those who commit it just before will have access to SNAP once they're released. 

 

Jennifer [00:17:44] Yeah, so like the precise timing of this law essentially randomizes you into the treatment or control group in lab experiment lingo. So as long as nothing else is changing at that date, and as you said earlier, no one kind of changed their behavior in anticipation of the law change or something. Everyone just behaved as usual, then they should look very similar. And it's kind of an ideal set up for testing this kind of program. So what data are you using for this project? 

 

Cody [00:18:13] I use publicly available data from the Florida Department of Corrections. They call it the offender based information system. It contains data on all people who were convicted and sentenced to incarceration in a Florida state prison. As long as they were released after October 1st, 1997. That release timing restriction ends up working well with the timing of the policy because almost all people convicted of drug trafficking are sentenced to two years or more in prison, so every one of interest for this policy will meet that October 1997 restriction. This data includes consistent inmate ID for each person that links all of their prison stays, the offenses associated with each stay and the offense dates. It also includes the person's name, exact date of birth, the sentence associated with each stay, the county they're convicted in and several demographic characteristics like age, sex and race. 

 

Jennifer [00:19:01] All right, let's talk about the results. What do you find is the effect of the SNAP ban for drug traffickers on their likelihood of re-offending? 

 

Cody [00:19:08] I find that the people who are subject to the SNAP ban are about nine percentage points more likely to return to a Florida prison than people who aren't subject to the ban. I want to stress, though, that the confidence interval in that estimate is about 1.7 percentage points to 17 percentage points. That's because by focusing on one state and on the people who commit drug trafficking right around August 23rd, 1996, I'm left with a relatively small sample of about a 1,000 people. So while I can reject the null that the SNAP ban has no effect on recidivism, I'm not able to produce a really precise range of possible effect sizes. 

 

Jennifer [00:19:43] And with that caveat in mind, how big is this effect relative to what we would have expected without the ban? 

 

Cody [00:19:49] Yeah. Based on the point estimate, the effect implies the SNAP ban increases recidivism by about 60 percent. That is admittedly very big. That's partly why I do emphasize that the confidence interval is anywhere from 1.7 to 17 percentage points, so anywhere from a 10 percent to a 100 percent increase. At the same time, I do want to highlight the context here. These folks are facing a really difficult time coming out of prison, and SNAP benefits are a meaningful source of financial assistance. And again, people convicted of drug trafficking may have particularly strong attachment to illegal labor markets. 

 

Cody [00:20:20] Also, it's not uncommon for work on recidivism or crime to find the policies can have fairly large effects. For example, that paper by Jillian Carr and Analisa Packham finds that the staggered dispersal of SNAP benefits reduces theft in grocery stores by about 33 percent. Crystal Yang's paper on the ban finds that the full removal of the ban reduces one year recidivism by 13 percent. She argues in her paper that that estimate is probably biased downward because it's not possible to measure a month of release in her data. So basically, in that data, if someone is released in 1998, they'll be flagged as subject to the ban. But some people released in 1998 will have entered prison before 1996. I think it's also worth noting that that 13 percent effect is averaging over people convicted of drug possession, lower level selling and trafficking. So it's possible that the effects are particularly large for people who are engaged in trafficking because their attachment to the illegal labor market might be a lot stronger than someone convicted of drug possession. 

 

Jennifer [00:21:14] And what types of crimes are driving this increase in recidivism? 

 

Cody [00:21:17] So after showing this main effect on overall recidivism, I then split all offenses into two categories that I call financially motivated or not financially motivated, and I test whether the ban increases the probability someone returns to prison for committing a new, financially motivated crime, or whether it increases the probability they go back because they committed a new non-financially motivated crime. The idea here is that if the effect of the ban is driven by loss transfer income, then we should see a larger increase in recidivism for crimes that are financially motivated. And that's what I find, I find an increase in people going back to prison because they committed crimes like selling drugs or theft, but no change in non financially motivated recidivism, so things like assault. I want to stress that this is slightly different than property crime versus non property crime, and I think this distinction more generally is important. A lot of papers have this same idea of testing the income mechanism by looking separately at property and violent crime. For example, a drug selling would not be classified as property crime, but it's very likely done for the express purpose of making money. And vandalism is classified as a property crime, but it's probably not so much about making money. Likewise, robbery is typically classified as a violent crime, but is seemingly motivated by income concerns as well. So that's my extremely mild take for this conversation is that if you're going to do an exercise like this and if your data allow it, then consider splitting offenses into different categories than just property or non property crimes. 

 

Jennifer [00:22:39] It's an excellent flag. So SNAP often requires that participants are working or engaged in some sort of job training. And as you mentioned, since people with criminal records often have a hard time finding jobs, this could make it more difficult for them to be eligible for this benefit in the first place. But it turns out that during certain periods, this work requirement for SNAP was relaxed. So tell us how you use that information in your analysis and what you find. 

 

Cody [00:23:05] That's right, and this is an important point. One of the big changes of the 1996 welfare reform was imposing work requirements for SNAP. In particular, people who the government calls able bodied adults without dependents, which is abbreviated as ABAWDs, are limited to only three months of snap receipt every three years unless they're either working 20 or more hours per week or participating in an employment and training program for that amount of time. For people with dependents, the requirements are a lot less stringent. But what's interesting is that at least in Florida, these ABAWD requirements are waived a lot. First of all, the requirements are waived nationwide from 2001 to 2003 and from 2009 to 2016. Second, these requirements can be waived at the county level if the county has an especially high unemployment rate or if it applies for another type of exemption. 

 

Cody [00:23:52] I got data on these waivers from Florida's Department of Children and Families, and a lot of counties actually obtain individual waivers like this. In 1998, 2000 and 2004, over half of Florida's counties had a waiver for the ABAWD work requirements. In 2006, it's closer to one third of counties, but again in 2008, it's well over half of counties. So while these work requirements exist, they're frequently waived in practice. I then use that data and the fact that I observe the individual's county of conviction for new offenses, and I find that people who are banned from SNAP are more likely to go back to prison, specifically because they've committed a new offense in a time and place where an ABAWD waiver was in effect. I think this makes sense because when an ABAWD neighborhood waiver is in effect, that's when SNAP is most generous for people who have access to it. And that's when we expect to see the biggest differences between those who are banned and those who have access. Of course, it's hard to attribute this solely to the effect of the waiver because, like I said, counties tend to get waivers when their unemployment rates are high, but it's more suggestive evidence about the role of SNAP here, and I think it helps square people's valid concerns about SNAP's work requirements in this setting. 

 

Jennifer [00:24:55] Yeah, that was super interesting and hadn't heard that about these policies before and haven't seen it studied before, so that was a neat insight. 

 

Cody [00:25:02] Thanks.

 

Jennifer [00:25:04] So you run a bunch of additional checks, including placebo tests, that is test that should turn up, null results if your identification strategy is isolating the causal effect of the SNAP ban on drug traffickers. So tell us about some of your favorite checks and why they're helpful. 

 

Cody [00:25:20] Sure. This goes back to our conversation earlier. But the key assumption of this research design is that people who commit drug trafficking right before August 23rd are similar to those who commit it just after. One major concern is that offenders, police, prosecutors, etc. know about this ban and are responding to it in some way. So one way to test whether that happens is to test whether there are any observable differences in characteristics and offenders committing trafficking before versus after the policy cutoff date. I can do that for all of the different characteristics that I see in the data. But a quick way to summarize that exercise is to take all of those characteristics and predict someone's risk of recidivism and then test whether offenders committing trafficking before August 23rd are observably different on this index measure of risk from offenders committing trafficking after August 23rd. Ultimately, I don't find any differences, which suggests that people who commit trafficking right before are quite similar to those who committed just after. 

 

Cody [00:26:14] Another way to test this is to see if the overall number of people committing trafficking changes. We see a drop in the overall number of trafficking offenses right after the ban, and maybe offenders have wisened up to the policy or maybe prosecutors have changed their charging policies. I test that in Florida and don't see a change in the number of offenses after the policy cut off. I think my favorite test along these lines, though, is looking at people who commit lower level drug selling offenses because these offenses aren't made exempt from the ban until May of 1997. So suppose people are keenly aware of this ban, and around August 1996, people who are committing these lower level selling offenses should also expect to be banned from SNAP upon release and anyone they interact with in the justice system should expect the same thing. So if the increase in recidivism for drug traffickers is driven by offenders sorting around the cut off in some unobservable way, we should expect people who commit lower level selling offenses to be sorted in the same way. And that would imply we should see a change in recidivism for that group, too. When I look at people committing lower level selling offenses before versus after August 1996, I don't see any change in recidivism among that group. What all of this means is that it seems like the experiment is valid in this setting - that we're comparing to groups who are otherwise pretty similar, except one is banned from SNAP upon release and one is not. 

 

Cody [00:27:29] Aside from this concern about people reacting to the ban, we also worry that perhaps there's some other state policy change that happens around August 23rd, 1996. If that's the case, we should see an increase in recidivism for people committing other crimes around that time. I look at people committing non-drug offenses or drug possession and I don't see any change in recidivism among those groups. Of course, another possibility is that given the seasonal nature of crime, perhaps there's something specific about drug trafficking offenses that occur after August. If that's the case, we should see an increase in recidivism for drug trafficking in all years after August 23rd. I look at all years after 1996, and in general, those estimates are centered around zero, and the estimate around August 23rd, 1996 is larger than the estimates around all other August 23rd. What all of that means is that the increase in recidivism after August 23rd, 1996 is specific to people committing drug trafficking and specific to the year in 1996, which is exactly what we should expect to see if it's caused by the SNAP ban. 

 

Jennifer [00:28:28] In the end, you use your estimates to calculate the social cost of this SNAP ban. How do you approach this and what do you find? 

 

Cody [00:28:35] That's right. Towards the end of the paper, I do this kind of rudimentary, cost-benefit analysis of the ban. Of course, to do something like this, you have to make a lot of assumptions. In terms of social cost, I make some assumptions that I think are very conservative. I assume that only cost of increased recidivism are the direct care cost of incarcerating someone, which the DOJ estimates to be around $10,000 per year and the cost of someone being a victim of whatever new crime the person recidivate commits. This ignores the cost of lost current and future wages for the individual who is actually reincarcerated, which we think might be upwards of 50k and ignores the potential physical or psychological cost of incarceration. On the other hand, it also ignores any potential deterrence effect of the policy and the possibility that it encourages legal work as well as illegal work. But when I do this, I find that the average cost of the ban is about $3,800 per ban offender, or over $70 million to date in Florida. 

 

Jennifer [00:29:30] And how does that cost compare to the money the state saved by not paying SNAP benefits to this group? 

 

Cody [00:29:36] Yeah, one thing that is missing above is the amount of SNAP benefits the state would have had to pay if it didn't ban these people. I don't include that as a benefit of the ban because it's a transfer and a transfer is one person's cost is another person's benefit. But I think people and policymakers are typically interested in the cost of these SNAP benefits. Here you have to make some assumptions, too. If I assume that on average, every person banned from SNAP would have received SNAP for one year at $150 per month had they not been banned, then the state saves about $1,800 per banned offender on average. That 1,800 saved is outweighed by the 3,800 in costs, so even if you completely ignore the benefit that people convicted of drug trafficking would get from receiving SNAP and completely ignore the costs they face from being reincarcerated, you still find that the cost of the ban outweigh whatever savings there are. If you don't ignore those things, the cost of the band is much, much higher. 

 

Jennifer [00:30:28] So that's your paper. Have any other papers related to this topic come out since you started working on the study? You mentioned a couple that were kind of contemporaneous, but others in addition to those? 

 

Cody [00:30:38] Yeah, I think there has been some other interesting work studying the effect of various social programs or financial resources on recidivism and crime in general. In terms of recidivism specifically, there was a paper that I think was also featured on this podcast by Caroline Palmer and coauthors published right around the same time as mine that showed that access to emergency short term financial assistance reduces recidivism. Since then, there's also been some cool work by Erkmen Aslim and coauthors that show state medicaid expansions led to a reduction in recidivism rates, and an interesting paper by Abhay Aneja and Carlos Avenancio-Leon, which shows that incarceration reduces access to credit upon release and that reduced access to credit then leads to an increase in recidivism. So those are a few examples of related work that I've seen since I started working on the topic. There's even more work if you include papers that look at the effect of social programs on crime rather than recidivism, and I think that area is really interesting and important, too. 

 

Jennifer [00:31:35] So what are the policy implications here, both of your results and the other work in this area? Policymakers that are listening, what should they be taking away from all this? 

 

Cody [00:31:43] I think the clearest takeaway from this paper and from other work is that the SNAP bans are probably not the best idea. They increase recidivism, which is very costly. Some states have done a full repeal already, but a lot of states did exactly what Florida did, which is repeal the ban for people convicted of drug possession, but leave it in place for people convicted of drug trafficking. I think the results of this paper and others suggest policymakers should consider just fully repealing the ban at this point. It seems like it was introduced to the bill solely for the purpose of being punitive. And this paper shows that that punitiveness is coming at a fairly big cost. Finally, I think there are two broader takeaways. First, when adopting a policy that might affect work incentives, policymakers really need to consider what type of work they are encouraging and consider the possibility that by taking away or reducing transfer of benefits, they might encourage people to increase work in the illegal labor market. Second, when we're looking for ways to reduce recidivism, increasing financial assistance seems to work pretty well. We see that in this paper, but we see it in other recent work as well. It seems like providing people with SNAP benefits or financial assistance in general can ease reentry and reduce re-offending. 

 

Jennifer [00:32:49] And what's the research frontier? What are the next big questions in this area that you and others will be thinking about going forward? 

 

Cody [00:32:56] When I talk to people about the SNAP ban, there are two questions that some people seem to want answered. The first is whether it has a deterrent effect. So now that the ban has existed for several years, do people factor it into their decision of whether to commit a drug felony and on the margin, perhaps some people decide it's not worth it because they don't want to face this ban upon release. Personally, I think it's super unlikely that people who would not already be deterred by the possibility of years in prison are somehow now deterred if we add the SNAP ban on top. Second is whether the ban increases participation in the legal labor market among people who are banned from SNAP. I think it's possible we see effects here, but I would imagine they're small just because it's particularly hard for people with a criminal record to find legal work. But nevertheless, if we want a full accounting of the SNAP ban, I think those are the two key empirical questions left on the table. I also think it's important to understand how SNAP or financial assistance in general affects recidivism for people who are in prison for other types of crime, particularly people who may have less attachment to the illegal labor market. For example, how much does this matter for people convicted of drug possession or assault? 

 

Cody [00:33:58] Finally, I've always had a hard time reconciling why the studies from the 1980s find mixed effects of providing unemployment benefits, but work today finds fairly consistent positive effects of providing other types of financial assistance. I think there are some potential explanations. You know, those studies cover a very different time period in the history of incarceration in the US. The unemployment insurance programs they studied had very high benefit reduction rates, which might have really discouraged any effort to find legal work and providing cash directly could mean people spend it on things that are complements to crime. But I think it's important to reconcile this, particularly because we'd like to know what types of programs are most effective at reducing recidivism and whether simply providing direct cash assistance would work just as well or better. 

 

Jennifer [00:34:40] My guest today has been Cody Tuttle from Princeton University. Cody, thanks so much for talking with me. 

 

Cody [00:34:45] Happy to do it. Thanks again for inviting me on. 

 

Jennifer [00:34:53] 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 enjoyed the podcast, then please consider contributing via Patreon. You can find a link on our website. Our sound engineer is John Keur with production assistance from Haley Grieshaber. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening, and I'll talk to you in two weeks. 

 

Jennifer DoleacRegression-Discontinuity, Resources, Re-Entry