Probable Causation

View Original

Episode 34: David Phillips

Jennifer Doleac

David Phillips

David Phillips is an Associate Research Professor of Economics at the University of Notre Dame.

Date: August 18, 2020

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

A transcript of this episode is available here.

See this content in the original post

Episode Details:

In this episode, we discuss Prof. Phillips' work on how emergency financial assistance affects crime:

"Does emergency financial assistance reduce crime?" by Caroline Palmer, David C. Phillips, and James X. Sullivan.

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 [00:00:19] My guest this week is David Phillips. David is an Associate Research Professor of Economics at the University of Notre Dame. David, welcome to the show.

 

David [00:00:27] Thanks for having me.

 

Jennifer [00:00:28] Today, we're going to talk about your research on how emergency financial assistance affects criminal behavior. But before we dive into that, could you tell us about your research expertize and how you became interested in this topic?

 

David [00:00:41] Sure. So my focus is largely on poverty and anti-poverty programs in the United States. So I kind of backed into being interested in crime in the US as somebody interested in poverty. So originally, awhile back, I was working on my dissertation work at a nonprofit employment agency in Washington, D.C., and I was spending some time around there looking at issues related to employment. And they're helping people find jobs in the middle of the Great Recession. And it dawns on you at some point that every other person who walks in the door has a criminal record that affects their job search. I think oh, you know, maybe I should actually pay more attention to the criminal justice system. And this is right around that same time that I think a lot of people were like - people were starting to throw around the term mass incarceration and people were reading the new Jim Crow and this sort of thing was happening. And I think in economics, there was this broader appreciation that, oh, maybe we should be thinking about crime in criminal justice as a first order thing when we talk about poverty in the US.

 

David [00:01:31] And so that sort of planted a seed in my mind and then where this particular project then came up is - so I work for the Lab for Economic Opportunities, or LEO at the University of Notre Dame. And LEO had this ongoing partnership with folks doing emergency financial assistance in the city of Chicago that's focused - a lot of the research they've been doing so far has been focused largely around homelessness. We started thinking about to what extent there might be a link between these types of social assistance programs and crime outcomes, whether we could link into arrested and the sort of thing as an outcome and this particular project was born. And I think it really does come out of that sort of recognition that when you're thinking about poverty and anti-poverty programs in the United States, a lot of times if you look at the cost benefit analysis, like outcomes related to crime are going to dominate. Right, if you look at the Perry Preschool Project, like - which many of your listeners may be familiar with, that early childhood intervention that's - you've got a lot of people as one of the most cost effective interventions we can do. Most of the benefits come from crime reduction. And so this idea that, hey, we're looking at this homelessness program as emergency financial assistance program, should we look at not just things like shelter usage and so on as outcomes, but, hey, we should we should be paying attention to crime and how this is affecting crime in the city of Chicago and whether it has an impact there. So I was able to kind of jump into a project that Jim Sullivan and some other people at Notre Dame had already begun looking at other outcomes and joined the team as we sort of launched on thinking about how this type of assistance might affect crime and criminal behavior.

 

Jennifer [00:03:00] So your paper is titled "Does emergency financial assistance reduce crime?" and it's coauthored with Caroline Palmer and Jim Sullivan. So let's start with basics here. What does emergency financial assistance mean in this context?

 

David [00:03:12] Yeah, so this probably exists for you and for me and for the listeners in their community. Right. So most places, if you call 211 or 311, you'll get a call center that can direct you to emergency financial assistance. And actually, these places are getting deluged with calls right now because of the COVID situation and lots of people having lost income because this type of assistance really is designed for smoothing over shocks. Right. So it's designed for a situation where you've got somebody who's in a stable situation, they've got a job or they've got income from public benefits, they have a stable place to live, they're in a stable situation and then something happens. They lose a job, their hours get cut back, their public benefits get reduced for one reason or another unexpectedly, and it's a temporary thing. Right. I expect that I'm going to be recalled to my job or I expect that my hours are going to go back up or that my public benefits are going to get restored after this temporary sanction, and then this emergency financial assistance jumps in in that kind of situation to provide temporary assistance. Sometimes it's literally just providing somebody with cash. That's happening more often right now in the current COVID situation. In the context in Chicago, it's going to be most of the time basically rent paid to the landlords that this person owes back rent to their landlord and they're going to pay temporarily this person's rent until their income gets back to the place where sort of its permanent level.

 

Jennifer [00:04:30] So as economists, how might we expect such assistance to affect criminal behavior? What are the potential mechanisms here?

 

David [00:04:36] I think there's a few for this one - some of them are more sort of traditional economics of crime mechanisms and some of them might be a little bit new. So I think that the standard one that might come to the mind of a lot of folks who think of econ of crime would be like a sort of Becker opportunity cost type of world, right. Where this is essentially or somebody has just faced an income shock and we're going to compensate for that. And so we're basically just providing income to some person. And that's going to affect the trade offs that this person makes between engaging in criminal behavior that might be income generating theft or what have you versus either leisure or some market behavior. In particular, right, if you think leisure is a normal good people are - work less when they have - when their sort of income drops on the sky from them. You might expect that to happen for crime too, right. If I'm committing crime in order to raise income and all of a sudden now I have this income, I don't need to commit the crime quite as much anymore. And so you might expect things like property crime to go down when somebody has an income dropped on them because they don't need to do the crime anymore. So that's - so one possibility, is that kind of sort of Becker ish, rational opportunity cost type of story.

 

David [00:05:43] There are a couple of other stories that that are also out there. Right. You can tell the behavioral economics story. You know, this is sort of you can think of the type of idea that like —— think has ported this into economics. A lot of thinking about situations of scarcity and how that affects how people's cognition, and how they make decisions, how it makes them focus on the present, and have difficulty sort of making optimal tradeoffs between the situation I'm in right now and something that might be way off in the future. And desisting from violent crime might be exactly the kind of thing that's kind of hard to do when you're very focused on the present in this current situation, you're sort of caught up in that moment and the consequences of that violence are down the road and a little bit less immediate. If you think a violent crime as rather than being a rational decision as being sort of a somebody making an action that's inconsistent with what they would like their long term self like to do, then providing income and smoothing out that situation of scarcity could help. Right, you know, if somebody's lost their job, you fill in that gap by providing enough income for them to pay their rent for the month that might provide them the space to be able to sort of think through what they really want to do, avoid violent situations with family or housemates or neighbors or what have you, and sort of reduce some crime by just giving the person some more space.

 

David [00:06:58] The third potential mechanism that I think is useful for this particular paper, but also more generally when thinking about income is to think actually about housing. This paper's going to have some connection to homelessness and housing stability, in particular because the financial assistance is going to be used mostly for paying rent, but that's actually also true for income in general. The biggest expenditure for most households is their housing, particularly for low income households and so housing itself shapes the environment in which somebody might take different actions about criminal behavior. Right. It shapes who I live with, it shapes how crowded that situation is. There's a lot of good examples. For instance, people might be familiar with Matt Desmond's book "Evicted" from a few years ago. But there's lots of stories from that book about people moving into really suboptimal housing situations, moving in with unstable roommates or very crowded housing or what have you because of lack of income leads to poor housing and then that, just like physical environment, creates interpersonal reactions that might lead to crime or violence or what have you. And so I think we're sort of imagining that there may be this sort of classic Becker mechanisms. There may be this sort of behavioral scarcity mindset type of mechanism and then there also might be sort of housing specific things that are going to matter in this context and more generally for situations where you've got low income folks with widely varying income over time.

 

Jennifer [00:08:19] All right. So before you all wrote this paper, what did we know about the effect of financial assistance on crime in practice? What research was out there?

 

David [00:08:29] So, I mean, there's a huge amount of research on employment in crime. Right, and that's a very broad literature that's out there. And I think we'd make the argument in this paper that that's - that is really important and really interesting, but it's a little bit different than understanding how income and income shocks and smoothing out those income shocks affects people's criminal behavior. Right. Because a job is income, but it's a bunch of other things, too. It's time use. Right. It takes me from maybe sitting on the couch to - and having nothing to do to being in a job. Right. It also had - there's some identity elements of being employed. So we think employment matters and it could matter for income. But we'd want to know a little bit more specifically about income and how people respond to shocks.

 

David [00:09:14] Similarly, like we have some literature on how people in criminal behavior and crime responds to shocks in general. Right. So there's the one of my favorites, the old card and doll paper on the National Football League and how in a particular city, when the home team loses unexpectedly, you see an increase in domestic violence, which is one of these terrible aha papers, at least for me, it's like, oh, wow. Yeah, that's, yeah that's going on around me all the time, that when we experience situations of shock that can lead to - that can lead to violence.

 

David [00:09:46] Specifically on income, I think the closest set of papers though really is about public benefits. There's this series of papers about public benefit cycles through the month. The original one that I know of is a paper by Fritz Foley from a few years ago, about a decade ago now. That was the first, I think, to document this, that there's a lot of things that follow this sort of food stamp cycle where people get their benefits at one point in the month, and then spending and a bunch of other things sort of happen concentrated around that point. Supermarkets are booming around that time in a lot of low income neighborhoods because people just got their benefits renewed. Foley's paper is one of the first to identify, hey, crime actually follows this pattern to. You see, when people's benefits are starting to run out at the end of their benefit month, that they start committing more crime and that tells you a couple of things, right? It tells you, one, that people are credit constrained in some way - or the constrain in some way that's not allowing them to sort of smooth out their income, that people's income is volatile enough that when they get the public benefit matters. And two that, it also matters for crime. Right, that like how much criminal behavior people engage in, depends on what resources are available to them at that point in time. And so I think that paper is one that's really informative about sort of laying the groundwork for saying like yeah, when - the timing in the volatility of people's income, even in an expected way, right something that people know about ahead of time, is - it causes problems for people's ability to cope with that and then this is from from criminal behavior.

 

David [00:11:15] I think you had Jillian Carr on the show. So people familiar from a couple of podcasts ago with her paper looking at trying to get rid of that cycle, basically. And so I think her paper - that one's really sort of contemporaneous with our paper. She was presenting that around the same time that we were presenting the paper I'm talking about. But I think it gets at a similar point right. Of hers - for those who aren't familiar, is looking at the situation in Indiana and Illinois, where they went from providing food stamp benefits all on the same day to spreading it out throughout the month so that different people got their benefits on different days. And you see a benefit of that, of sort of disrupting this like monthly cycle of crime. It's caused by people's lack of resources at particular times of the month. So I think that's where we're sort of jumping in is in a same - similar space to where Jillian and Analisa were jumping in saying, like, okay, we have this evidence that the resources that people have available to them could affect crime. Now, what happens when we actually respond to that? What if we try to actually smooth that thing out - smooth income out? How is that going to affect the crime that we see happening around us?

 

Jennifer [00:12:20] So, as you said, these emergency financial assistance programs exist in cities across the country. So why don't we know more than we do about what the impacts of these kinds of programs are? What are the research hurdles that you had to overcome in order to measure the causal effects of this kind of program on criminal behavior?

 

David [00:12:37] I mean, some of it's just a data issue, right. So some of it is - because you really need to observe a set of people who have experienced these shocks and that's kind of hard, actually, right, to screen on people who have lost a job or had their benefits cut. That sort of a data barrier to begin with. Let's identify people who've had a shock and then let's also see a group of people then who are receiving some intervention. Right and so you might not be able to pair those together. Right. That like - you might be able to go to something like unemployment insurance benefits and see somebody that experienced a shock, but then to match that up with, oh, here's a person who got some response to that situation - that they received some assistance, matching that up is kind of hard. And then you need to link those two pieces of information, right this person had a shock and they received some assistance up to an outcome on crime, which probably then resides with a third different agency. And so there's part of it, just the data linking difficulty of connecting all of those pieces of information is challenging. And so for us, we had to sort of overcome that by linking data from a nonprofit organization that runs homelessness services to the police department records to be able to overcome that. And that's  kind of a challenging thing to do. So I think one challenge we had to overcome was, yeah, just actually getting all of the data together in one place.

 

David [00:13:47] I think there's also like an identification issue. So there's a - doing cause and effect here is particularly hard because you're - we're particularly interested in people who've experienced income shocks. We want to look at a group of people to the reason that they're getting this assistance is because things just went south. And in a lot of program evaluation situations, we're really worried about the Ashenfelter Dip. We're worried that people who sign up for a program are signing up for it because they're in a bad situation. So they're going to look like they're doing worse when they sign up for the program, but that's not because the program. Right. It's because there was something else that caused them to sign up here. Here, the situation is actually a little bit worse because we're actually directly selecting on the Ashenfelter Dip. We're saying to be eligible to get this assistance, you have to have had something bad happen to you. So forming a comparison group that has experienced that same shock but not receive the assistance is going to be particularly hard to do, and particularly important to do to make sure that we don't mistakenly look at the situation and say, wow, these people are doing so terribly that they got the assistance is like, well, that's why they showed up in the first place.

 

Jennifer [00:14:47] Right. And then on top of that, you also have - these are people that are selected on having experienced a shock, but are also motivated enough or knowledgeable enough to know to reach out to this organization, which kind of is an additional layer of, like, selection that might be a problem here, right?

 

David [00:15:03] Yeah. So you get, yeah - and that could make it go the other direction. Right. So then if you're able to account for the fact that people have experienced these shocks but not their ability to actually access services or be aware of services, then you could get the other direction. Right. Then you could say, wow, these people are doing so well. And in reality, it's just that they had the motivation to deal with it or the connections or the resources to deal with it. And it's those things that matter rather than this assistance.

 

Jennifer [00:15:23] Right. So in this paper, you consider the effects of emergency financial assistance to people who call Chicago's Homelessness Prevention Call Center, which I'll refer to is because HPCC because it is shorter. So what does the HPCC do and who is eligible to receive assistance from them?

 

David [00:15:41] Yeah, so they're essentially the call center. So Catholic Charities Chicago runs this call center. And so they're basically on the other end of 311. You call 311, you say I'm behind on my rent, I need some help. They forward you to the HPCC, and they've got staff there at the phone bank who are helping to identify whether the people who call are eligible. So most people who call are not eligible actually for the program. They're actually looking for a pretty narrow group of folks who have experienced these sort of temporary shocks. So they screen on four dimensions is the way that they describe it. So one of them is self-sufficiency. So it has to be the case that this person has enough permanent income to cover their rent. If somebody has - his rent is too high relative to their sort of long term income, what they expect their income to be in the future, they're not going to help them with this assistance, they need something more permanent than this sort of temporary assistance.

 

David [00:16:29] It has to be the case of the person has had an eligible crisis. So they have to have had some event. So they have to be able to point to I got laid off, here's my public benefit cut, my public benefits were at this level, and my SSI check got cut by this much and to be able to document that. So it has to be the shock, right. It has to be something happened, and it has to be temporary. Beyond that, it does have a connection specific to housing. They also screen on people being imminent risk for homelessness. So think of that as having an eviction notice. So these are people who are currently housed but who very soon are not going to be housed if this thing doesn't get fixed. And finally, they only take people whose situation is solvable. So if somebody owes their landlord six months of back rent and several thousand dollars, and that's more than the cap for any fund for which they're eligible, then they're going to get turned away, too. So you get this group of people who have experienced shocks and that shock threatens their housing. But this intervention looks like it could fill the gap. That's what they screen on.

 

Jennifer [00:17:24] And how much money are people typically getting?

 

David [00:17:26] They're usually getting about a month's rent. So this is Chicago approximately a decade ago. So think - I think the average for our sample is like $800 a month, which - so to some folks is going to sound like absurdly small if you're living in like Seattle or something like this. But that's about what - that would be a little bit low for now in Chicago, too, but is about one month's rent. And the actual amounts for these programs varies a lot across the country. But the idea of filling in one month's rent is very common. So you might have a program in Seattle that pays a couple thousand dollars, and one in Chicago that only pays $1,000 because they're trying to do the same thing in theory, even though the amounts are different.

 

David [00:18:04] One thing that is important to note is among homelessness and housing programs, this is much cheaper than most other homelessness programs. So that's expensive, right, relative to some other things we might imagining doing it that are much lighter touch. But among housing programs that - compared to like a housing voucher that might permanently pay most of somebody's rent and costs, you know, ten thousand dollars a year or something. This is for housing intervention pretty inexpensive.

 

Jennifer [00:18:30] Yeah, that's a great point. And yeah, I mean, as you said, this is in some ways trying to solve a different problem - this temporary shock to permanent income. But if this works, then it is a really cheap intervention relative to other options.

 

David [00:18:44] Yeah, I think that's right. Because, I mean, the tradeoff of something like this is you're testing a much broader net, right? Because there's a bunch of people who are going to be able to resolve their situation on their own and it's hard to tell who those people are and only focus on the people at the margins who this makes the difference. Right. So it's casting this big net. It's not as well targeted as, you know, giving a housing voucher to somebody who's already homeless, but it's less expensive per person that you serve. And so there's this tradeoff between poor targeting and inexpensive per person that if you manage that correctly, it might be really cost effective thing or it might be a big waste of money if we're just giving this to people who would be able to figure things out themselves.

 

Jennifer [00:19:20] Yeah, so, just thinking terms of the eligible population, who are these people in general? Like do you have other kinds of descriptive stats at your fingertips about what group winds up being eligible for this program.

 

David [00:19:33] Yes, I mean, they're quite disadvantaged for instance. The screening on imminent risk of homelessness. Right, is going to mean you have a very disadvantaged group of people. So that that's going to show up in a bunch of different ways, whether you look at people's pre existing income levels or ethnicity, where they live in the city of Chicago, and so on. Now they are going to be people who have some attachment to income. Right. So they're not going to be below zero income. There's gonna be a lot of public benefits income, but also a lot of employment, perhaps more than other similarly disadvantaged people. For an economics of crime crowd, right, probably one of the unique things about this group is that it's largely female. So these are heads of households who are seeking homelessness assistance. Homelessness prevention works not exclusively, but the vast majority of people who are receiving homelessness prevention funds across the US are going to be female headed households. And so that's - so when we're looking at crime for them, we're looking at criminal behavior of women, which is a little bit, as I'm sure your listeners know, most crime is committed by men, and so that makes that that a little bit different group. It's interesting to note, though, that actually if we look at pre people's arrests prior to when they show up to the program, even though it's a largely female population, there still are more likely to be arrested in the average person in the city of Chicago. So it's still a group of people who had a lot of contact with the criminal justice system, even though most of them are women.

 

Jennifer [00:20:53] Yeah, that's interesting. Okay, so it turns out conveniently for you that it is unpredictable when money will be available for people who call in and this creates a natural experiment. So talk us through this process and how it gives you something akin to random variation in who gets the financial assistance that they need.

 

David [00:21:11] That's right. So they - so the Homelessness Prevention Call Center doesn't flip coins with people. So they're not running a randomized controlled trial, but the process they had in place already sort of approximates one through their existing process, which allows us to evaluate this. So one of the key things that happens right - so they're getting calls that are coming in from people calling information and they get fed into this call center. And so there are a couple of aspects that are important.

 

David [00:21:35] One is that the HPCC operators screen people on eligibility first up front. And so they do sort of a long intake with folks that gather some information on them, like identifiers that we can use to match the arrest data like people's names. They also screen on some of those broad eligibility criteria I mentioned before, self-sufficiency, having a crisis, et cetera, and also information about what the need is that the person has. So lots of information about the situation that the person is in is collected for everybody, whether or not they end up being eligible, whether or not they end up actually getting funds. And so we see these selection criteria for everybody and we can narrow down the eligible population from this much larger haystack of people who call and turn out to not be eligible.

 

David [00:22:19] Now, once you have that eligible group, right, the ideal from the scientific point of view would be then that we flip a coin and give assistance if it's heads and don't give assistance if it's tails, and they don't do that. But they do something pretty close to that. What they do for the eligible folks is at any given point in time, the HPCC has a list of different funds that they can refer people to. So they're essentially like this central distribution point where all of these different agencies, some of them government, some of them nonprofit, take all of their clients as referrals from the HPCC and so because it's this sort of central distribution point, they're sort of then passing people off to all these different - these different agencies. But because there's this large number of agencies and the situation of each of those agencies is fluctuating over time, any given agency happens to have space for somebody varies a lot from day to day.

 

David [00:23:09] So, for instance, some organizations are getting funding from the state, and Illinois was going through various cycles of budget crisis during this time and so you have some of these things shutting on and off. You've got nonprofit organizations that are staff constrained, and they've got a certain number of slots each day, so maybe this I've got five slots and tomorrow I've got 12. And so whenever a given person calls HPCC, and they find out, hey, this person is eligible, they start going through that list of potential places they can refer them to. And sometimes there's a place that's got a spot for them and sometimes even though they're eligible and they're just as good of a case as anybody else and just as good of a fit for the program, they just don't have space. And it really varies wildly from day to day in an unpredictable way.

 

David [00:23:49] I think one of the things that early on that I have to give credit to Jim Sullivan's coauthor on this, identifying this earlier on in the previous paper, even in there, like the manual for the people who run this call center, there's this really interesting set of text where they talk about, like, what do you - what do you do if somebody - when you tell somebody no, and they ask you what should I do? Like when when will the fund be available? And they literally tell them, tell them I don't know because it varies so much. When should I call back or should I call back? Well, that's up to you. And if they ask what's the best time to call, say, I don't know. There's no best time to call like the need is so high, we don't have enough for everybody, which is right - which is like tragic for the operators and for the people in that situation, but from a research point of view is kind of perfect because it's really unpredictable. People have a hard time knowing if you call in a given day whether you're going to be able to get assistance or not. And that sort of gives us the purchase to be able to have a treatment group, a comparison group who is quite similar, but doesn't get the assistance, but for whom we have the same data.

 

Jennifer [00:24:46] Right. And so, yea so just step us through how exactly you're using that variation to then measure the causal effect of financial assistance on criminal behavior.

 

David [00:24:55] Yeah, so we take the whole haystack of callers. We narrow it down to people who are eligible based off of having had a crisis and so on. We also narrow it down to people who are first time callers because some people do call back multiple times and that in of itself is quite self-selected. It might get into sort of motivation concern you mentioned before. We still keep the first call for everybody, but we don't want to examine three repeat calls as a separate sort of instance, so we limit down to a group of people who are sort of calling for the first time who are eligible. And we're going to compare people who are referred to funds when they're called, when they call first, people are not referred to another agency for funds and so that would be like a simple difference in means. Right we could compare crime outcomes for both those two groups, just like an RCT, and say that's a treatment effect. It's a little bit more complicated in this context because the sort of agencies that they're referring to, some of them have their own eligibility requirements. So Agency A might only accept veterans or Agency B might only pay back rent, but they won't pay security deposits because that's what some people need, some agencies won't do that. And so there is - there sort of conditional random assignment here.

 

David [00:26:00] Right. So there's some need to control for differences between different people who show up with somewhat different situations that are eligible for a larger versus smaller proportion of these other funds and so we end up doing that as well. So in the end, what we're doing is we're running a regression of some crime outcome on an indicator for whether the person gets funding, but then controlling for these factors that affect eligibility for the individual specific funds. But the nice thing is, again, it's a centralized intake and so we actually observe all of those things. So for each of those agencies, we have the sheet that says here are the things that make you eligible and we can say, okay, here - these different need requests of rent versus security deposits or whatever, these are the categories that matter and we could control for these things.

 

Jennifer [00:26:44] Yeah, it is a really cool data set. And I'm sure lots of listeners are already thinking about how they would love to get their hands on this data set. So talk about the data that you're using for the project as well as the outcome measures that you're particularly interested in here.

 

David [00:26:58] So the sample is defined and based on information and all of these control variables and treatment stats - all that is defined for the Homelessness Prevention Call Center data. So yeah, the sort of richness of that data is one of the things that makes this possible. Right, it overcomes this problem of identifying a comparison group, controlling for potentially confounding variables, and being able to link the existence of a shock and what that shock was, because that's in the data, that's part of the eligibility - did you have a crisis and what kind of crisis? Being able to match the shock to whether the person got assistance. So like most of the data problems are resolved just by the call center. And so the Lab for Economic Opportunities, where I work, LEO has had connections since its founding with Catholic Charities. Right. We're at Notre Dame, a premier Catholic University, Catholic Charities, a premier Catholic Social Service provider and so we work with a bunch of different folks, in a bunch of different contexts, but one of our big partners is - are different branches of Catholic Charities. And so there's a partnership between Catholic Charities of Chicago, which runs the call center, and the University of Notre Dame that allows us to share that data.

 

David [00:27:58] And then the second piece of it then was linking it to crime. Right, cause the call center has no information on people's criminal behavior after the call, and so that was a - the next step that we had to manage. And so there we worked together with folks in the Chicago Police Department and Chapin Hall, which has lots - at the University of Chicago, which has lots of connections in terms of data within the city of Chicago to connect those two data sets, so that we would be able to observe whether people who call the call center were arrested at some point before or after the call. Now, arrests aren't a perfect outcome we can talk about that a little bit if you want, but it's different from committing crime. It's just crime that's observed and then leads to an arrest and you can have both type one and type two errors. You can have people arrested who didn't actually commit a crime and plenty people who commit a crime, they don't actually get arrested. But it gives us some indication of engagement and criminal behavior that we can then use as an outcome.

 

Jennifer [00:28:52] All right, well so, let's talk about the results. What do you find is the effect of receiving emergency financial assistance on subsequent arrests?

 

David [00:29:00] Sure. So overall, the probability of arrest falls a bit when people get emergency financial assistance relative to when they don't. It is a little bit noisy. So we sort of intentionally in the paper don't try to pin too much on one number for the overall drop in arrests. But it does seem like the - in the immediate aftermath of the call that people are less likely to be picked up by the Chicago Police Department when they get this type of assistance.

 

Jennifer [00:29:26] And then do those effects vary across different types of crime?

 

David [00:29:29] Yeah, and that's one of the big things that I think we discovered in this paper that was perhaps a little bit surprising to us was the pattern that showed up here. What we ended up seeing is actually a big drop in violent crime, but not property crime. So you see violent crime arrests dropped by about half and so, right violent crimes arrest aren't super common in general. And that's true in this group as well. So it's something like 2% of our sample gets arrested for a violent crime in the year after they call for the people who don't get funds and that drops to about 1% for people who do get funds. It's a small change in terms of probability, but it's a big change in terms of the prevalence of violent crime and given how socially costly violent crime is, that's - I think in lot of ways that's the headline result of the paper is that you can move something like violent crime in a very noticeable way with this relatively modest assistance. Again, like 98% of people in the sample aren't going to be arrested for violent crime anyway, but for that couple of percent that we're going to be engaged with the Chicago Police Department on something related to violent crime, if you can cut that in half, that's a - that's a tremendous success. So I think that's the big biggest headline result from a cost benefit and is this emergency financial assistance making a difference point of view.

 

David [00:30:37] From an economist point of view, the other thing that we find that's - I think matters less for cost benefit, but is sort of intriguing from thinking about why does crime happen point of view is that we actually see property crime go up. So we see an increase in larceny for - going up by about the same amount. So three years later, there's not much difference in sort of overall number of arrests and it's because the initial drop in violent crime is offset by some increase in property crime that sort of slowly picks up over time and happens a little bit later on. Property crimes a lot more common in general so sort of a change in how much of property crime is contributing to it is a little bit smaller because it's just a bigger base to start from. But there is sort of this noticeable increase in property crime that happens after a little while.

 

Jennifer [00:31:22] Yeah, to talk more about that, I think you say in the paper it was about - at about the 12 month mark.

 

David [00:31:27] Yeah. So the - on the property crime side, the interesting thing that we see is and this was sort of an unexpected, I should say. Right, because like the Becker story goes the exact opposite direction. The Becker story is - or sort of, you know, maybe extended Becker story it is - yeah, you provide people with some income. They don't need to commit crime anymore and so you'd expect, especially property crime, to go down. Right. People don't need to steal anymore. And the type of thing that we're seeing here really is that type of crime right. It's larceny committed by female heads of household, primarily. They're getting arrested in commercial establishments. I mean, it looks a lot like shoplifting. Right. And so, you know, why is shoplifting going up when you give people financial assistance? And as you said, it really it doesn't - it kind of delays a little bit and it sort of picks up at 12 months. And we puzzled over this for a long time. And the thing that helps sort of crack an understanding what was going on there for us was splitting out who this was happening for by the type of assistance they were receiving.

 

David [00:32:30] So most people are getting assistance from the HPCC for back rent. So they're behind on rent with their landlord. They're about to get evicted. And now we're going to pay a month of rent to your landlord so that they don't kick you out. Some people are instead getting security deposit assistance. Things are bad enough with their landlord that it's going to be hard to resolve that situation. So we're going to move you into a new place and pay for your security deposit instead. Well, what we see is - for particularly that group of people - the people who are getting security deposits we see them 12 months later having this uptick in shoplifting. And can you nail down exactly what that is? No, but but it suggests a story, right? It suggests a story that, you know, most families that's a good situation, have the security deposit paid for, but there's some families that are getting this new place set up for them and 12 months later, they find themselves in the same situation. They're sort of over their head in terms of finances and shoplifting as a way to make up some income to be able to resolve that situation. And so you have sort of maybe this little bit of an unintended consequence of getting financial assistance, leading to somebody getting into a worse situation, which then leads to some increase in property crime later on.

 

David [00:33:30] Now, I do want to emphasize the main original point, which is that that's like from a public cost benefit point of view, that if we can prevent one assault and at the same time cause one instance of shoplifting to happen, we want to make that trade off. And so I don't think that's a knock on the program itself, but it does tell us something interesting, I think, about how we run these programs and how we screen that - the sort of screening on is this a sustainable situation, particularly if you're paying for a new rental contract for somebody, you don't want that rental contract to come up 12 months later and have this person be in this tough situation where I really have to pay that off so I can stay in my place and forcing them into a difficult situation that leads to shoplifting.

 

Jennifer [00:34:07] Right. It suggests that even if the current program is cost effective, you can imagine another version that's even more cost effective if we can figure out how to solve that problem.

 

David [00:34:15] Yeah, so like, I mean, I think the screening side and the targeting side of like, who do you actually provide this assistance to. The more you sort of think about these programs and this type of assistance, that question not just of is it effective but for whom is it effective and for whom is what type of it affected is a really important part of really designing these and making them as beneficial as possible.

 

Jennifer [00:34:37] Well, so those results say a little bit about mechanisms. I know you talk more about that in the paper. So what else are you able to say about the mechanisms that are driving this overall effect?

 

David [00:34:49] Yes, so on the violent crime side, I think there's a couple of different things that are - that can be going on. So first some things that we think that it's not right. So one sort of like more complicated version of a sort of rational criminal behavior, Becker-ish type of story could be that - you could be actually shifting people from violent property crime, if there's sort of like a index of like severity of crime, we're moving people from more severe to less severe crimes. That doesn't seem to be happening. It turns out that the groups of people are different, so that all of the property crime increases coming from families and all the violent crime decreases coming from single individuals. So it's not sort of like - we're not just, like, shifting group one from one to another. So that's not why violence is going down. It seems to us like there's a couple of different things going on, one that we can sort of directly show and one that we hypothesize, but it's hard to sort of get direct evidence on.

 

David [00:35:36] One, I think is - I do think housing is part of the story. We do see, for instance, the same group of people, singles having less arrests for homelessness, specific crimes. So things like trespassing, so people getting arrested for trespassing outside, that goes down too. And so you see some indications - and the other paper that Jim and others wrote that demonstrated this program does reduce entries into homeless shelters and you see some sort of similar things if you look at just like address changes. So decreasing housing instability, preventing people from ending up on the streets or in shelters or probably more likely for more people like doubled up with family and friends and really crowded housing situations, like we see evidence that that's happening. And it's not a hard line to draw from that to decreases in violence. Right. There's a lot of qualitative evidence, Desmond's book and others that right - if you put a bunch of people into a small space, that you just end up with more violence either between people and their housemates or it spills into the neighborhood around or what have you. And so it's the idea that housing and particularly crowded housing and unstable housing may lead directly to violence, I think is something that we can trace out and see evidence that the mechanisms for those are active in the data by seeing things like those arrests for trespassing.

 

David [00:36:49] Another thing that we think is probably going on but is a little bit harder to test with the data that we have, is that some of this - the sort of behavioral story of Mullainathan and company, right that this is just lowering people's level of stress in the emergency situation and allowing them to process the situation that they're in. Right. We're talking about people who are in a crisis situation, who were already low income, have lost their job, and now have an eviction notice, right, providing space in that context seems likely to me that that allows people to manage the process of avoiding violence better than if they don't have that income available. Now, that's one that we can't test. But I - if I could test, I think I suspect is part of the mechanism what's going on is creating that sort of space for people to be able to manage their very difficult situation better decreases violence.

 

Jennifer [00:37:42] I wanted to highlight one other result, because I'm particularly interested in prisoner reentry and was excited to see this paper in that context. As we're thinking about targeting and who might benefit the most from programs like this, I think you find that people who have at least one prior arrest benefit a lot. Am I remembering that correctly?

 

David [00:38:00] That's right. And so we do see bigger effects for people who have some history with the criminal justice system, right, because we see arrests for like the five years before the person calls and so we can split up the sample into people who've had an arrest in the past and people who haven't. Now, it's - even though we have a group of people who's more likely to have been arrested by the Chicago police than the average person, that's still a relatively uncommon thing. So we find evidence that, yeah, this makes a big difference for people with an arrest history. I do think it opens up an interesting next question, which is - which we can't really answer in this paper. But I think it's an important one for the future, which is what is financial assistance look like for people who have more severe contact with criminal justice system? What about a population of folks who, you know, is sort of more typical group of folks who it would be a high risk of violent crime, a group that might include a lot more men, right, it might include people with much more extensive histories with the criminal justice system. Now there's other challenges to doing financial assistance in that context for why it isn't as - done as much, but this at least suggests that the benefits could be pretty big if you see the same sort of mechanisms there. So it's - we can't say a lot about it because other than in our sample, you have people with the longer histories do have a bigger reduction in future crime, but I think it's like this little tidbit that points to like, wow, it would be really good to know more about this.

 

Jennifer [00:39:20] Yeah, I totally agree. So as you discussed in the paper, you might worry that the people who receive assistance are simply easier to find because they're more likely to be living in stable housing or they weren't kicked out when they got that eviction notice. So that could mechanically lead to more arrests even if actual criminal behavior isn't changing. So, the police are able to find you if they want to arrest you, basically. So how do you convince yourself that that isn't driving the effects that you see here?

 

David [00:39:46] Yes, we convinced ourselves on this, and we tried to convince referees as well - I suppose we did at least well enough to get it published - and I'm interested whether actually from like sort of a broad criminal justice research audience, how connvincing people find this, because I think we found it quite convincing. But I think there was you know, there was a range of opinions on this.

 

David [00:40:04] The way that we looked at it was one thing that you see in the Chicago arrest data is you did see warrants in our data, which are largely - and actually through this project now I'm doing some more work on pre-trial, and failure to appear, and that sort of stuff, so this is a lot more familiar to me now, actually, than it was when we were writing this paper, but most of these warrants are for people not showing up at court. Right. They don't - they don't show up - failure to appear rates are quite high or at least if you've seen others on the show talking about FTAs right it can be super high and is a big issue in the criminal justice system.

 

David [00:40:36] For our purposes the fact that these are in the data can actually be a little bit helpful. Somebody who, right, somebody calls in on a particular day, right, and they're calling the call center and they get the assistance or they don't. If they get arrested the next day on a bench warrant, that's almost certainly for a case that happened before the call. And so there's at least this initial window after the time of the call where arrests for bench warrants are really picking up, just can the police find you. They're not picking up new criminal behavior because you haven't had time for that to happen, really, but you've had - but you have had time over that window for the police to try to find you on this existing bench warrant.

 

David [00:41:11] And so what we do is we look at that a bit and we don't see any effect on arrests for bench warrants, and we take that as some evidence that this isn't just about them finding people more. Now, that's, you know, there's standard errors around that and there's some uncertainty about that, but at least when we looked for it in the way that we could, we didn't find any evidence of that. The other thing we were concerned about, too, which is related is that our data are limited to the city of Chicago and you might be worried that if people have housing instability, that they that they move to another jurisdiction and out of the data set. And there, I think we were a little bit less concerned about that issue than the first one, because migration routes out of Chicago are pretty low for this population. You're going to have a lot of migration within Cook County, but not a lot of people actually leaving the city, among groups who are - you know, folks who are living on the south side of Chicago.

 

Jennifer [00:41:58] So the study was published in the Journal of Public Economics last year in 2019. Are there other studies that have come out recently that add to our understanding of the effects of financial assistance on crime?

 

David [00:42:10] True. So I think there's a couple of different areas where we've seen progress. I mean, I think the - well starting with, right I mentioned Jillian Carr and Analisa Packham's paper before on cycles and food stamps disbursement and sort of spreading those out around the month, and how that might affect crime I think so - yeah, so, I mean, their paper came out about the same time. So I think that's  one thing that we sort of learned alongside each other. They have continued down the path and are continuing to produce really interesting papers about that particular topic. The original paper was more on property crime and now this more recent one is about domestic violence and they actually find something that's more in the opposite direction. So that I'm still puzzling over a little bit is they - when Indiana and Illinois sort of smooth out this public benefit cycle, they're seeing increases in domestic violence. And their argument is that this is sort of like a within household resource allocation issue, that you're actually - that the public benefit itself is a source of conflict in the household and that messing with that might lead to increases in domestic violence. And so we have we're finding something different in Chicago, and so I think reconciling those is interesting. But I don't know if it's - or the fact that the HPCC they're directly paying the landlord, maybe that makes a big difference. Maybe, right, the lesson here is to - if you're working with particularly female heavy population of public benefit recipients who might have concerns about domestic violence, that you want to pay directly to the source rather than giving them agency over it, not because you want to give them agency, but precisely because you're worried about somebody else turning that into a violent situation as this conflict of the resources, I don't know. So I think their paper is really interesting and one that's that's complicated my understanding of this a bit that I'm still wrestling with.

 

David [00:43:49] Another set of papers that I think is related, that's been interesting, is there's a series of papers on fracking and crime that is - right there's this - there's a big literature on employment in the fracking literature, so Brittany Street has a paper, —— paper from a couple years ago, Emily Owens and the coauthor has a paper on fracking and crime as well. There's like this combination of both the employment and the income effects happening for different groups of people in the Dakotas and elsewhere where there's been an increase in fracking and so I think it's sort of an interesting natural experiment to see some of this apart and so I think that's another set of papers that I found to be quite interesting.

 

David [00:44:26] I think we're actually going to see a lot more specifically - so none of those are specifically on emergency financial assistance, right, they're not on these type of like HPCC type programs. I think partially that's because a lot of the research on these sort of programs that are like the Homelessness Prevention Call Center, the research is in progress right now. I think in the coming years, both because of COVID, right, there's a - emergency financial assistance has exploded in the past couple of months all around the country because there's so many people experiencing these shocks right now, and there's so many places that are lottering off the assistance or doing it in other ways that lead to credible research designs. So I think we're going to see a bunch of papers on how emergency financial assistance matters in the COVID environment. And I think there were also just - LEO and other folks who are also working on trying to bridge to the next step of doing emergency financial assistance and studying it in the context of something that's more explicitly a randomized controlled trial. But a lot of that stuff is so - like sort of long term, that it's a big lag. So I think there's a bunch of process right now.

 

Jennifer [00:45:25] That's great. I'm looking forward to seeing all those papers.

 

David [00:45:28] Me too.

 

Jennifer [00:45:30] So what are the policy implications here? What should policymakers and practitioners take away from this paper, along with all the other studies you just mentioned?

 

David [00:45:39] A couple of things. I mean, I think one - like the fact that we're talking a lot about emergency financial assistance during the COVID crisis, I think is right. Right. I think that a lot of  - the federal government has done a good bit and lots of local governments and local agencies have been trying to fill in the gaps on that. Even if targeting is bad on things like these. Right. Even if only 2% of people are going to commit a violent crime in the absence of this thing or be arrested for it or in - only a couple of percent are going to end up in a homeless shelter. Right. There's - if you're affecting these really costly outcomes for a small number of -small percentage of people, it can still be - you can still pass a cost benefit test, and that's what shows up here. So I think emergency financial assistance in a broad way is something that we're getting increasing evidence that it makes sense even when the targeting is hard to do. So I think that's one thing to learn.

 

David [00:46:25] I think a second thing to learn about is that like more broadly for social assistance, often we think about this as from sort of like what's the justification for these things? It's well I'm sort of sitting behind this Rawlsian veil at the beginning of time, and we're making society, and I don't know who I'm going to be, so we should, you know, I want some insurance against being in the worst situation, and so we should sort of build in this social insurance. But there's also an externality argument for this, right. I mean, there are situations where poverty and income volatility leads to outcomes that are bad for everyone, like crime. And so if we see changes in - well things that are essentially public goods in response to some of this type of social assistance, that gives a different type of argument for social assistance in general and helping people smooth these out because it doesn't just affect them. It also affects the people around them.

 

David [00:47:13] I think the third thing - and I've been thinking about this a lot in the context of COVID is that, there's a lot of violent crime that - our paper suggests that a lot of violent crime is sort of a result of just like poverty and stress spilling over out into the world, and that you can actually produce a decent amount of sort of like garden variety assaults and maybe not the most severe things like murders and such, but like sort of like the bulk of - like the most common violent crimes a lot of that is coming from these sort of day to day interactions, things related to housing and such. And so when you think about the current situation that we're in right now, where lots of people have had the shocks, particularly then people are also like sort of confined to being at home, which might be a good situation or might be a really bad situation for some people. I get really worried about what's going to be happening in terms of domestic violence and so on in the current context and how can we respond in a way that mitigates some of those effects that we could try to get rid - you know, can we try to get rid of some of these potential violence spikes in violence through things like financial assistance?

 

Jennifer [00:48:13] So Emily Leslie and Riley Wilson have a new working paper where they try to measure - quantify the effect of the lockdown on domestic violence you were just talking about.

 

David [00:48:23] Oh, yeah. I saw that.

 

Jennifer [00:48:23] And they are finding that, so it turns out they're like strong seasonal trends in domestic violence calls, but once you control that - for all that, it does look like domestic violence calls have gone up during this period, which is what everyone was worried about. I think when people have been talking about that, they've really been focused on telling everyone they have to be home, even if home isn't very safe. And you're pointing to sort of a different mechanism there that it's really like the stress - you know, that can be a problem, too, but layering on the stress of losing your job and losing your income is a problem that we could find a way to solve, at least in this current moment.

 

David [00:49:01] Yeah, exactly. I mean, it may be the case that everybody has to be at home, but if everybody can be at home in a situation where they don't - they're not wondering how in the world we're going to pay the rent once the eviction moratorium goes away. Right. That's a less volatile environment than one in which I'm just - the only thing that we're doing is that I'm helped, and I'm a little bit less stressed.

 

Jennifer [00:49:18] Exactly. So what's the research frontier? What are the next big questions in this area that you and others are going to be thinking about going forward?

 

David [00:49:26] So we've touched on a couple of these. I mean, I think some of them are the usual research design things. And so I think there is scope for, right this is not a random assignment study. There's, you know, if we wanted to talk for the next two hours about potential things we could poke at in the research design we could do. Right cause it's not perfect. And so actually doing a random assignment study of something like this, I think is something that's needed and there's external validity questions, right. Housing is very different and income volatility is very different in Chicago than it is in a lot of other places around the country and so picking up additional pieces of evidence about this or more generally, I think is important.

 

David [00:50:01] The current situation that we're in right now is different than, you know, the environment that we were studying, which was happening a couple of years after the Great Recession, right so that's both unstable, but for very different reasons. So I think some of it's the usual sort of internal validity, external validity stuff.

 

David [00:50:16] Beyond that, like sort of big questions. I do think this question of like what would this look like for people with much more severe criminal histories is a really interesting one. It's a lot more challenging because like the risks of it going wrong and having unintended negative adverse consequences might be higher. Right. If - and so we have that sometimes income - positive income shocks are associated with mortality risk, drug use, and so on. Like you might be worried about some of those a bit more, but you also might potentially have much bigger benefits. Right. So I think that's one piece of - I think that's part of the frontier, and I think that a piece of the frontier is really the current COVID environment too - that emergency financial assistance is as important now as it's ever been because so many people are experiencing the shocks, understanding how this stuff works right now it's really important.

 

Jennifer [00:51:03] My guest today has been David Phillips from the University of Notre Dame. David, thanks so much for doing this.

 

David [00:51:08] Thank you so much.

 

Jennifer [00:51:14] 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 in our website. Our sound engineer is Jon Keur with production assistance from Elizabeth Pancotti. 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.