Episode 65: Felipe Goncalves

 

Felipe Goncalves

Felipe Goncalves is an Assistant Professor of Economics at UCLA.

Date: January 18, 2022

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

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Prof. Goncalves's work on whether police enforcement and crime:

“Do Police Make Too Many Arrests? The Effect of Enforcement Pullbacks on Crime” by Sungwoo Cho, Felipe Goncalves, and Emily Weisburst.

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. My guest this week is Felipe Gonsalves. Felipe is an assistant professor of economics at UCLA. Felipe, welcome to the show. 

 

Felipe [00:00:26]  Hi Jen, thanks for having me. 

 

Jennifer [00:00:27] Today, we're going to talk about your research on police enforcement and crime. But before we dove into that, could you tell us about your research, expertize and how you became interested in this topic? 

 

Felipe [00:00:38] Sure. So I would say, broadly speaking, that my research is applying tools from labor economics and public economics to questions in crime and policing. So I work in two main areas within crime and policing. So the first one is on understanding police behavior. So the behavior of police officers. In that area I've worked on studies on how do you measure racial discrimination by police officers? How do you use data on police civilian interactions to measure any racial discrimination? I've done a study on the impact of police unions on officer misconduct and another study on the impact of police overtime pay on their arrest behavior. So broadly speaking, these are studies using tools from labor economics to study police officers as workers and understand their behavior. 

 

Felipe [00:01:36] The other area I work on, which we'll talk in much more detail about today, is understanding offender behavior. And in particular, these studies are asking what law enforcement tools actually have an impact on offending behavior. So looking at various types of law enforcement interventions, law enforcement actions and what effect do those interventions have on crime and offending? 

 

Felipe [00:02:01] And so I became interested in this topic early on in graduate school because I was more generally interested in studying racial inequality. So I had been interested in a long time in studying racial discrimination across various settings and my interest in crime and policing grew out of that background. So this was in the early 2010s when there was a growing public attention towards the problems of the criminal justice system. So that was the origin of my interest was just seeing this growing public focus on this issue, thinking about how I could apply, what I was learning in my classes to this topic and I found that it was a good match of the tools I was learning in labor to this topic. But I also found as I was working on it, that I ended up liking some of the more mundane specifics of research on this topic. So I like learning about the institutional details of policing, learning about the specifics of various departments. I like the part of the research process where you have to go out and find the right setting and data to answer your question. So that's what kept me working on this topic. But that was really the origin of my interest in studying crime and policing. 

 

Jennifer [00:03:16] The paper we're going to talk about today is titled "Do Police Make Too Many Arrests? The Effect of Enforcement Pullbacks on Crime" and it's coauthored with Seung-Hui Cho and Emily Weisburst and in this paper, you consider whether reductions in police arrest activity affect local crime rates. Set the stage for us what do we know about how police affect crime more broadly? 

 

Felipe [00:03:39] OK, so obviously there's a very large literature on policing and criminology more generally. So this is not a new area, and I would say that there's a very large literature on some specific parts of policing and how they impact crime. 

 

Felipe [00:03:55] So there's a very large literature on the impact of police employment on crime and within that area, there are a bunch of empirical studies that use a variety of methods to ask what happens to the crime rate in a city when that city's police department hires an additional police officer. So these studies will find some credible research designed to compare a police department that hires one additional officer to some similar department that didn't hire an additional officer. And these studies as a whole find fairly robust evidence that the marginal hiring of a police officer does reduce crime, meaning at current levels of police employment in the average U.S. police department, if you hired one additional police officer, there would be a reduction in the crime rate. And so I'll mention one paper in this literature to give you an idea of what kind of empirical methods these studies are using. So a frequent coauthor of mine, Steve Mello, he has a paper looking at the cops grant program from the late 2000s. So this program was a federal grant program that was giving grants to local police departments to hire more officers, and he looks at departments that apply for this grant money and get the grant money with similar departments that apply for this grant money and don't get the grant money. And what he finds is that those departments that do in the grant money they do go on and hire more police officers and relative to the losing departments, they also experience reductions in crime in the years after. And so he can use that to say, OK, well, what was the impact of these additional officers on the crime rate? So that's one broad area where we know fairly convincingly that police employment does reduce crime even at the margins at current levels of police employment. 

 

Felipe [00:05:51] There's also a lot of research on the impact of police presence on crime. So this is asking beyond just the number of officers that are employed by a department. What is the impact of actually placing an officer on a specific street corner on the amount of crime on that street? So to give you an example of a study in that area, there is a study by Draca, Machin and Witt looking at the 2005 Central London terrorist attacks. So this is a terrorist attack in London that led to a significant increase in the number of police officers stationed around these very dense locations within the city of London. And so what they find is that after this terrorist attack and after the increased presence of police officers, there's a big reduction in the number of reported crimes around those areas. And interestingly, what they also find is that a few weeks later, when the department reduces the number of police officers in these specific locations, crime jumps back up to the pre terrorist attack levels pretty much instantaneously. So right along with the reduction in police presence suggesting that the presence of police officers is also playing a role in the very local levels of crime. So I would say that those are two areas where we know a lot about what the impact of police is on crime. Obviously, there's a much broader literature in policing that covers other dimensions, but I would say that these are the two areas where we have a lot of abundant and really credible research that policing does have an impact on crime. 

 

Jennifer [00:07:29] Yeah. And so your goal in this paper is basically to peer inside the black box of policing a little bit. So to start thinking about what police are actually doing with their time, which has been much more difficult to study and in particular, you're interested in whether changing the number of arrests officers make has an impact. So why might police arrest activity affect crime? 

 

Felipe [00:07:51] Ok, so the way I see it, you can think about there being three primary ways in which arrests could impact crime. So the most direct and mechanical way that you could imagine arrests impacting crime is that, well, when you arrest somebody that often will lead to jail time or prison time for the arrested person so that physically incapacitates the individual. And if that person was going to commit some crime, if they were not jailed or incarcerated, now they can't commit that crime. So that's going to directly reduce their short run ability to commit crimes. So that's broadly people think of that as incapacitation. So arrests can incapacitate people from committing other crimes. There could also be what's called the specific deterrence effect. So this is where individuals are deterred from committing crimes after being arrested simply from the experience of being arrested. So this could be because now that you're arrested, you realize that the likelihood of getting caught from committing an offense is higher than you expected. It could be that the process of being arrested and jailed is worse than you imagined, so that experience was worse than you imagined. So these things could deter people from committing crimes from the actual experience of being arrested. Now, I'll note here that this impact doesn't actually have to be crime reducing the experience of being arrested may actually harden and individual and actually increase their future offending. So maybe you realized that being arrested is not as bad as you thought it would be. And I'll mention here some of your research on misdemeanor prosecution that your study with Amanda Egan and Anna Harvey, where you find that being charged with a misdemeanor offense actually increases future offending. 

 

Felipe [00:09:37] So that's a different but very closely related type of punishment. And I mentioned that to say the impact of arrests and crime doesn't necessarily have to be a negative effect. The final kind of catchall way that arrests can impact crime is what's called general deterrence. So this is where individuals are deterred from offending from the overall realization that they're now more likely to be arrested if they commit a crime. So somebody is looking out in the world trying to decide whether or not to commit an offense, and they have some sense of what's the likelihood of them getting caught? If police increase arrests, maybe people realize that the likelihood of getting caught goes up and that might deter them from crime. So those are the various ways you could imagine that arrest activity could have an impact on the crime rate. 

 

Jennifer [00:10:20] Yes and so in this paper and I guess and studies in general, in many cases, the best hope is to get the net of all those effects together. And many people listening might think, you know, this seems like a really basic question how do we not know the answer to this question about how police arrest activity affects crime? So why is this question so difficult to answer? What were the hurdles that you and your coauthors had to overcome when you started thinking about how to measure the causal effect of arrest activity on crime? 

 

Felipe [00:10:47] I agree that it's definitely a fundamental question just to give you one fact here. Police departments in the U.S. arrest over 10 million people a year, so they make over 10 million arrests a year. So I agree that it's very important that we understand what the causal effects are these enforcement actions. And then it's perhaps surprising that we don't really have a great idea of what the effect of these different arrests are. 

 

Felipe [00:11:12] So the challenge was answering this question of what's the impact of arrests on crime is that arrests and crime are often moving together over time for reasons that are unrelated to causality. So there are lots of reasons why arrest and crime can be very over time and across departments. That has nothing to do with arrests go up and it has some effect on crime. So the first one is that if crime goes up, you should perhaps mechanically expect there to be more arrests because there are more arrestable offenses out there in the world. So there are more opportunities for police departments and police officers to make arrests. You could also imagine that police departments may change their enforcement strategies when crime is high. So when crime goes up, police departments might respond and say, OK, well, we have to change our arrest activity to help reduce crime. So to give you one big example of this, police departments in the early 1990s began to adopt a more aggressive policy of increasing the number of arrests against low level offenses, but this was done specifically as a response to heightened crime at the time. So if you looked at this period, you may see that places with higher crime also had a greater number of arrests, but you're actually kind of looking at a reverse causality here that the crime increase is what's driving the increase in arrest activity. The final challenge, I'll note here, is that both arrests and crime could be impacted by some third confounding factor, so arrests and crime could both be driven by some other change in the criminal environment. So as an example of this, consider what has happened recently after several high profile deaths at the hands of police. 

 

Felipe [00:13:02] So we've had lots of these high profile events in the last decade where there's often social unrest and public outrage in reaction to these deaths at the hands of police. Simultaneously with this public outrage there is often also an increase in the number of reported crimes in the city where these incidents happen and you also observe that police are reducing their enforcement activity, potentially as a reaction to the heightened scrutiny of the departments of police have some pullback in their activity. This is what's been called the "Ferguson effect." There's been a temptation by some commentators on these events to say that the decline in enforcement is actually what's causing the increase in crime. You say, Well, look, police are pulling back their enforcement activity, we're also seeing an increase in crime that's why we need all this enforcement activity because we see that crime is going up. But I would argue that it's very hard to know in these settings what's the impact of the arrest decline on the crime increase because the entire environment has changes, the entire social and criminal environment is changing. It's really hard to isolate the arrest decline component and its contribution to the crime increase. So these are various reasons why it's quite difficult to estimate the impact of arrest activity on crime separate from these other features of the criminal environment. 

 

Jennifer [00:14:27] You're going to use officer deaths as a natural experiment in your paper. Officer deaths created enforcement pullbacks that you argue only affect police behavior, not civilian behavior. So tell us about these events. 

 

Felipe [00:14:40] Right. So the events we are studying are line of duty deaths of police officers. So as I just noted, when you have these high profile deaths by police officers, there are a lot of things that are changing about the environment. 

 

Felipe [00:14:55] These are really high profile events. We argue that Line of Duty deaths of police officers are also going to be events that change police behavior, but it's more plausible that other features of the environment are not changing as much, and in particular, these are events that don't have as much social unrest related to them. So these cases that we're studying, this line of duty deaths are cases where officers are dying while working on the job. There are also a few cases here that we're going to be looking at where the officer is actually off duty, but is acting in their capacity as a police officer in some kind of dangerous incident. But broadly speaking, you can think of these as cases where officers are working on duty. So these incidents typically occur during some kind of police encounter, like a traffic stop and the majority of these cases, the officers being killed by gunfire. So it's some kind of interaction that escalates and the individual kills the police officer, typically by gunfire, though there are some cases we see here where it's by vehicular assault. So these events are very salient events, as you might imagine for the police officer's peers and their departments. So these are highly traumatic events that impact the officers who are working with the officer who's been killed. And I'll note in particular, one researcher who's done a lot of work on this Michael Sierra Arevalo, who is a sociologist at UT Austin, has done a lot of ethnographic research on police culture and how they memorialize these events and one of the big takeaways from his research is that these Line of Duty deaths are a very important part of how police officers view their job and how they view their jobs and life or death terms. 

 

Felipe [00:16:48] So on the one hand, these are very salient and important events on the police side. What we argue is that on the civilian side and on the general criminal environment side, these events are comparatively much less noticed relative to death at the hands of police and so one thing we've done here is we've looked at Google search trends for the names of these police officers and compare them to what happened to Google search trends of individuals who die at the hands of police after those events. And what we find is that, as you might imagine, there are a lot of searches for individuals who die at the hands of police and there's a kind of a sustained attention in these events for several weeks. In contrast, you don't really find much attention broadly when there's an officer death in the line of duty. There is a little bit of an effect. There are people searching for this officer's name, but that effect dissipates pretty quickly. And so what we argue is that this is a, from a research perspective, an ideal setting for studying what happens when police respond to an event, but the environment as a whole is not really changing. 

 

Jennifer [00:18:00] Yeah. And so what you're going to find is that these officers pull back and arrest fewer people. So is the idea here that the event that the death of their colleague just makes it more salient? How dangerous their job is? Is that the general idea? 

 

Felipe [00:18:13] Yes, that's the main idea. So what we're going to end up finding, like you said, is that police respond to these events by reducing their enforcement activity. And what we interpret this is that police are responding to this event with heightened fear about the dangers of their job. 

 

Felipe [00:18:32] And so one piece of evidence related to this is that if you look across departments and how they respond to these events, the biggest declines in enforcement activity are actually coming from smaller departments where these events are more infrequent and perhaps more shocking. So that's consistent with the idea that these officers are more traumatized or more surprised by these events and have a bigger response to the event. And so we're arguing that these are events that are being caused by changes in fear and some alternative hypotheses that you could consider here is that we don't think that this is due to some change in police strategy. We don't think this is due to some change in the number of officers working out on the field. So this is holding fixed the number of officers who are working out in the field holding fixed the employment of the department. There's a reduction in the enforcement activity of officers. So that's why we think this is a good setting to isolate the impact of enforcement activity, holding fixed these other dimensions of policing that we discussed earlier that there's a lot of research showing that those other dimensions matter. 

 

Jennifer [00:19:42] OK, so then tell us in a little bit more detail exactly how you use the timing of these police deaths to measure the causal effects of police arrest activity on crime. 

 

Felipe [00:19:51] OK, so in the simplest terms, what we do here is we're looking at basically the time path of enforcement activity around the dates of these police deaths. So when we look at is how many arrests are being made by a police department before one of their officers dies in the line of duty? And then compare that to how many arrests they're making right after? That's the main gist of the empirical strategy. 

 

Felipe [00:20:18] There are some more complicated details here, which is that, well, there might be features of the environment that are changing over time arrest rates and crime rates are changing over time and maybe you want to account for that. So what we do is a difference in differences design. We're looking at the time path of arrests of these departments that experience a line of duty death with similar departments that don't have a line of duty death at that time. So we argue you can use those departments as a credible comparison group for the time path of arrests activity. That's the intuition of our empirical strategy, and you can think of that as a difference in differences. So it's the difference in arrests before and after this line of duty death for the effect the department compared to the before and after for some similar department that doesn't have a line of duty death on that day. 

 

Jennifer [00:21:13] Great. Yes, very intuitive and also a kind of strategy that's also very visual so it would be great if we could show graphs, but that is the downside of podcasts. So imagine graphs in your mind, listeners. OK, and then the other thing you needed once you had this great idea for a natural experiment and the idea for how you would analyze the data, you needed the actual data. And this paper involved what I would call a Herculean data gathering process. It's truly amazing the data that you all put together. So tell us about the data you have and how you gathered at all. 

 

Felipe [00:21:47] Right. So we collected a bunch of different data sets to do this study. So the first dataset that we used here is the data to actually identify these events of officers dying in the line of duty. 

 

Felipe [00:22:00] So to do that, we are using data from the FBI that's called the law enforcement officers killed or assaulted dataset. So this is a data set local police departments report this to the FBI information on when officers are killed or assaulted in the line of duty. We took that data and linked it to a police maintained database called the Officer Down memorial page that records all of these events of officers dying in the line of duty. And so the events that we look at are the combination of these two data sets. So that's the data we use to identify these events. 

 

Felipe [00:22:36] The main outcome data that we're looking at is from the uniform crime reports data. So this is data that is provided from local police departments to the FBI, and it provides information on monthly arrests by departments and monthly measures of reported crimes. A lot of departments only record this information at the annual level we'll probably get into this maybe a little more, but there are definitely some issues with the uniform crime reports data in terms of the data quality of these data. So what we do is we restrict attention to departments that are reporting these data at the monthly level. That's the baseline data is the uniform crime reports, data for arrests and crime. In addition to these data, we've collected a bunch of additional data sets that help us get a better sense of what's going on during these Line of Duty deaths. So to explain some of these. So one of these data sets we use is data on traffic stops by police departments. So we got data from the Stanford Open Policing Project, which has done this amazing job of collecting micro data on traffic stops by a lot of police departments. 

 

Felipe [00:23:49] And we took that data to construct measures of the number of traffic stops being made by these police departments at the monthly level. 

 

Felipe [00:23:56] Another big data set that we're using for this project is data on 9-1-1 calls to police departments, so this was definitely the biggest list on our end in terms of data construction. So we made a series of public records requests to local police departments for their micro data on 9-1-1 calls. So we got this data for over 70 police departments, and we're using that to construct a measure of the calls to police departments at the monthly level, but and I know we'll get into this in more detail later, but we're also going to use this data to measure police reporting practices. So we'll be able to say, are there any changes in how police report crimes after these Line of Duty deaths? 

 

Jennifer [00:24:44] All right. So let's talk about the results. So first, you consider the effects of police deaths on arrest activity. What do you find? 

 

Felipe [00:24:51] So what we find is that there's a sharp decline in arrest activity by the affected department after these officer deaths. So across all arrest activities, we find about a seven percent decline in arrests in the first month after the incident, and we find about a four percent decline in the month after the incident. After that, there is a return to the baseline level of arrest activity before the officer death. So this is about a one to two month decline in officer arrest activity. These percentage declines are pretty similar across arrests types, so we see a seven percent decline in serious violent arrests and arrests for property crimes. 

 

Felipe [00:25:41] We also see about a seven percent decline in arrests for non index crimes. So these are crimes like DUIs, and we also see about a seven percent decline in arrests for what are called quality of life offenses. So these are things like arrests for disorderly conduct and loitering and you can think of these as the most discretionary arrests. So these are arrests that are typically initiated by police officers and are not initiated by civilian 9-1-1 calls. 

 

Felipe [00:26:13] So while we see a similar percentage decline across all of these categories, the biggest magnitude of decline is in these low level quality of life offenses. So the number of arrests that are now foregone that are not being made are much lower for these low level offenses. So that's where we see the biggest decline in arrests activity after a line of duty death. 

 

Jennifer [00:26:39] And then what is the effect of those police deaths on local crime rates? 

 

Felipe [00:26:44] So despite finding a decline in arrest activity for this two month period, we don't actually find evidence of a crime response. So the baseline finding here is that there is no change in crime after Line of Duty death and not only that, our estimates are fairly precise. And so we can rule out fairly small increases in crime rates. So to be a little more technical, here our 95 percent confidence intervals can rule out that crime went up by more than three percent. So to give you an idea of how surprising that result is, going back to the research that I cited at the beginning of police employment and crime, if you had had a seven percent decline in police employment, you would expect a more than three percent increase in crime. Based on this research, that's looked at in police employment. Similarly, if you had experienced a seven percent decline in police presence on street corners, you would also expect more than a three percent increase in crime. 

 

Felipe [00:27:56] And so we're we're ruling out here crime increases that you would maybe expect from these other dimensions of policing. So we think that this is a fairly interesting and novel finding because this is suggesting that the marginal impact of arrests is qualitatively different from these other dimensions of policing. 

 

Jennifer [00:28:17] And across places since you have so many places in your in your data set, the effects of these police deaths varied a bit. So you could look at whether the effects on crime depended on the intensity of the change or the length of the change in arrests. So what did you find there? 

 

Felipe [00:28:33] Right. So like I said, the average effect that we get after a line of duty death is a one to two month decline in activity and about a seven percent decline in arrests. But you might be curious about, well, what would be the effect if it was a more than seven percent decline or a longer term decline in arrests? And so we look at this directly by examining heterogeneity across departments in how big the arrest decline is and how sustained the arrest decline is. And surprisingly, what we find is that pretty much across all types of departments in terms of how big this arrest decline was, we don't really find crime increases. So if you specifically focus on departments that have very big magnitudes of arrest decline, we continue to find no crime increase. So another way of putting this is that we don't really find a threshold level of arrest decline beyond which crime now goes up. Similarly, for duration of arrest decline, we looked specifically at departments where their arrest activity goes down by more than four or five months and even focusing on those departments, we don't really see increases in crime. 

 

Felipe [00:29:53] I do want to add a caveat here, which is that when we actually look at heterogeneity across departments, this is certainly more suggestive evidence in the sense that we're just stratifying across the departments in terms of the effect that we find so that we don't have, we could call quasi experimental variation in the magnitude and duration of arrest declines, so I would say this is suggestive. 

 

Felipe [00:30:18] But what we're finding here is that there really isn't a threshold of magnitude or duration of arrest decline beyond which crime is going to go up. 

 

Jennifer [00:30:28] So interpreting this as the causal effect of arrest activity requires assuming that police deaths only affect crime rates through their effect on arrests. So that is, if officer deaths affect criminal behavior as well or how often crime is reported to police, the story here would be more complicated. So you consider this assumption in a few different ways in the paper. So tell us about the analysis you do. 

 

Felipe [00:30:51] OK, so to analyze these settings and be able to interpret what we find as a change of arrests on crime, we have to assume that the only feature of the environment that's directly responding to the officer death is the department's arrest activity, and this means assuming that other dimensions in the environment are not responding. And so you can think of this as broadly two things. So the first is that civilians are not responding to the officer death, and the other one is that other dimensions of policing are not changing. So to test the assumption that civilians are directly not responding to the officer line of duty death. What we did is that we looked at crime in cities where there was no arrest response. So we have a subset of cities where there's an officer line of duty death and there's no reduction in arrest activity. 

 

Felipe [00:31:50] And you would think that if civilians are directly responding to the line of duty death separate from arrest activity, then maybe you would see some change in crime. For those cities, and we end up finding is that there's no real change in crime in these cities where there's no arrest response. 

 

Felipe [00:32:09] The other thing you might be concerned about is other dimensions of enforcement changing after a line of duty death. And so two things you could consider here first may be police officers change, say the amount of use of force that they engage in because of heightened fear. And so to look at that, we examine deaths by police after these line of duty deaths, and we find that deaths by police actually don't change after these events. You might also be concerned about changes in police deployment, so maybe the number of officers working out in the field goes down. 

 

Felipe [00:32:48] So the other thing we'd be concerned about is that other dimensions of enforcement might change. So one example of this is that you might imagine that use of force by police might change after these Line of Duty deaths. So to look at that, we examined deaths by police after these line of Duty deaths. And what we find is that deaths by police don't actually change, which is suggestive evidence that the kind of force by officers is not changing. And the other concern you might have is that police deployment or allocations of police officers out on the street could change after Line of Duty that death. So one example of this could be that you might have some officers who go on leave to mourn or grieve the death of their fellow officer. You might have some officers that are assigned to go solve the murder of their fellow officer. 

 

Felipe [00:33:44] And so those might be situations where the number of officers out on the street changes. And so what we argue here is that the magnitude of decline that we find is big enough that it would be hard to explain that decline by a reduction in total number of officers out on the street. So for the typical department, our arrest decline is comparable to about twenty two officers arrest activity. So it would have to be that twenty two officers in the typical department are not working for two months after this line of duty death, which is fairly large to be what's driving this effect? 

 

Felipe [00:34:23] The other thing I'll note here is that we look directly at these other dimensions of enforcement activity in one police department. So we look at these other dimensions in Dallas, Texas, where we have lots of rich micro data on a use of force and police deployment. And what we find there is that there aren't really changes in police force activity, and there's no change in the number of officers working on the streets. So that could as direct evidence in one department that these other dimensions of enforcement are not changing after a line of duty death. 

 

Jennifer [00:34:57] You also consider the effects of police deaths on 9-1-1 calls using all those data collected so. First, tell us why you do this and then tell us what you find. 

 

Felipe [00:35:08] OK, so the reason we collected all this data on 9-1-1 calls is that crime report data, which is the data I've been discussing so far, is not only a function of crime happening and civilians reporting it to the police, it's also a function of police actually writing a crime report. 

 

Felipe [00:35:29] So a crime only happens in the uniform crime reports data if a police officer actually writes down a report, say a crime happened, one story you could imagine here is that after a death of a fellow officer-- police officers not only reduce their enforcement activity, but they actually reduce their propensity to write a report. So maybe you go to someone's house because they they called 9-1-1, but whereas before you would have said that a crime happened and you wrote a crime report you're now a little more hesitant of your activity and you don't write a report. And so that never even shows up in the crime data. 

 

Felipe [00:36:06] So if that's happening, we might be biased away from finding crime increases. So you can imagine a story where crime actually did go up, but police are reporting fewer crimes, and that's leading us to say that there's no change in crime. And so that's why we wanted to look at 9-1-1 calls, which are not filtered by police reporting activity. And what we end up finding is that similar to the crime report effects, there are no changes in 9-1-1 calls after a line of duty death. So we similarly get no change and you can similarly reject very small increases in the number of 9-1-1 calls. 

 

Felipe [00:36:47] What we can also do is actually directly look at reporting practices. So for many of these cities where we've collected 9-1-1 data, the records also record whether the officers wrote a report for individual incidences, whether an officer wrote a report for these individual incidents. And what we find is that the probability of writing a report doesn't change after a line of duty death. The other thing we can do with these data is look at a greater range of types of crime. So the types of crime that are reported to the uniform crime reports is a subset of types of crime, and it's typically more serious types of offenses. 

 

Felipe [00:37:28] 9-1-1 data records everything civilians are calling the police about, and we continue to find no change in the full set of crimes after a line of duty deaths. So that gives us greater confidence that we are not missing any change in crime in these kind of lower level types of offenses. 

 

Jennifer [00:37:48] And then finally, as you mentioned, you have really rich data from Dallas, Texas, and you use this to dig in a bit more on a specific event that happened in that city. So tell us a little bit more about that event and what you find in the Dallas data and what all of that analysis buys you. 

 

Felipe [00:38:05] So what we do in Dallas, Texas, is that we collected a bunch of micro data from their police department on all of these outcomes that we've just been discussing, but with much more detail so we can not only see crime rates. 9-1-1 calls arrests, but we can also see use of force by officers. We can also see who's out on the street working. So this is a setting where we can get a lot more detail on what happens after these Line of Duty deaths. 

 

Felipe [00:38:39] So what we do is we study a specific incident that happened in 2018, so Dallas Police Officer Rogelio Santander was working off duty and he was involved in an incident where there was a shoplifting at a store that he was working in, and he went to respond to the incident and he was killed by the individual. And so this is an officer that again, he was not explicitly working in the line of duty as a police officer, but he did engage in this incident in his capacity as a police officer, and he died in this incident. 

 

Felipe [00:39:15] And as we've seen in the nationwide analysis, there was a significant reduction in arrest activity by the Dallas Police Department. So there's this kind of heightened fear by Officer Santander's fellow officers after this incident, and we actually see a 30 percent reduction in arrest activity in the Dallas Police Department. So this is a big response to this incident and this decline lasts for about nine weeks. So this is a fairly significant sustained reduction in arrest activity. And what we find is similar to our nationwide analysis. 

 

Felipe [00:39:53] There's no increase in crime or increase in 9-1-1 calls after this incident. So that continues to confirm our baseline finding that arrest activity at the margin doesn't seem to be affecting crime or 9-1-1 calls. 

 

Felipe [00:40:11] The other thing that we find here is that use of force by Dallas police officers doesn't change, and the number of officers that are taking 9-1-1 calls, the number of officers that are out working patrol doesn't change after this incident. So again, this is further confirmation that number one, the criminal environment doesn't seem to be changing. Civilians don't seem to be directly responding to the incident. 

 

Felipe [00:40:37] But second, officers are responding by reducing their arrest activity. But these other dimensions of enforcement don't seem to be changing. 

 

Jennifer [00:40:45] So as we start to think about what to take away from all this one question people might have in mind is whether the marginal arrests that aren't being made due to an officer death. So the types of arrests that officers are not making during these pullbacks that you're examining are the same as the marginal arrest that would not be made if this were deliberate policy choice. So if if we told our police departments to make fewer arrests, are these the same types of arrests that they would not make in that case? What are your thoughts on this? 

 

Felipe [00:41:15] Yeah, that's a great question, and I think that's a difficult question to answer. 

 

Felipe [00:41:19] And I think it's going to depend on the policy that you have in mind. Let me give you an idea of one policy here where I think our study is particularly relevant. So police departments have a lot of discretion in their enforcement activity, even above and beyond what the criminal law is and what criminal law says is legal or illegal. And in the last several decades, there has been a proliferation of this idea of broken windows policing, which is a policing philosophy that covers a lot of different ideas, but one particular component of it is the idea that you want to have a lot of enforcement against low level offenses. And the idea there is that if you increase enforcement against low-level offenses, that's going to deter overall crime and we think of our study as being particularly informative for that. We argue that our study is suggesting that at the margin at current levels of enforcement, which includes quite high levels of arrests for low level offenses, you could reduce arrest activity at the margin for these low level offenses without sacrificing public safety. So if you reduced enforcement against low level offenses by, let's say, seven percent, you're not going to see significant increases in crime. 

 

Felipe [00:42:44] Now again, I do want to be careful in saying this because the variation that we find in this that we're looking at in our study is not a specific policy that says we're going to change arrest activity. It's looking at these events where an officer dies and then the police department responds with a reduction in arrests. And so it's not directly one policy that says, Well, let's reduce this type of arrest. It's a change in all types of arrests. But in a way, I actually think that that is very relevant for department initiated changes in enforcement because it's the officers that are choosing to reduce arrest activity. 

 

Felipe [00:43:21] This is a reduction in arrests that they have the discretion to say, Well, maybe these are not the important arrests to make, but I think is interesting here is that it suggests that maybe officers have a notion of what are the discretionary arrests and what are the ones where if you don't make them crime is not going to increase. And so I would say this is very informative for these more department initiated types of policies, things like broken windows policing. 

 

Felipe [00:43:46] Another question here is how informative is this for some type of policy that is not initiated at the police department level, say decriminalization policies? I would say you need a lot more research to actually say what the effect of those kinds of policies are. I think our study is informative for that. But it's certainly not the end of the line for answering this question. You want to have actual studies that look at specific decriminalization policies to know what kind of effect you're going to have in those types of events? 

 

Jennifer [00:44:17] Ultimately, we want to see policy changes of the type that you're discussing to know what the effect of those policy changes would be. 

 

Felipe [00:44:24] Yes, definitely. 

 

Jennifer [00:44:25] But your paper is obviously super interesting because people are discussing those kinds of policy changes and are worried about what the effects could be. So it's useful to use this as a similar example. So, yeah, I mean, I guess I've been trying to think through this to you because this feels like the million dollar question as we're trying to make this parallel. And it seems like there are two possible stories. One is that the cops are responding to these events, the deaths of their colleagues by saying, like, I'm not going to arrest anyone who scares me, it's not worth it. I'm just going to keep driving. If that's the case, then it's hard to imagine that a deliberate policy choice would say, don't arrest scary people. That's going to be less relevant. 

 

Jennifer [00:45:01] But if the story is more like the one? That you just told where officers are kind of have a sense of whether a particular arrest has a real public safety benefit and they can say yes, there's any more benefit to arresting this person than that person and now their sense of what the potential costs of the rest are has gone up. Then they're just drawing a line a little bit higher. And so they're pulling back on exactly those marginal arrests that are just less worthwhile, which seems plausible to me in this setting than, yeah, I think it is exactly the parallel that we would imagine if this were a deliberate policy choice. But it's super interesting to think through kind of what exactly are these officers thinking? And is it the kind of decision they'd be making in this different context? 

 

Felipe [00:45:51] Yeah, I think that's a good way of delineating it that you could imagine not only what I described above. They have a good sense of what the crime reducing benefits of each arrest would be, and they just say, Well, let's increase the threshold for making an arrest. And it just happens to be these marginal arrests that don't really have any benefit. The other story here is that maybe there's a divergence in incentives for the officer and for society, and there are these arrests that officers don't like making that are beneficial for society, like the arrests of some, you know, scary guy that actually is a danger to society. The officer says, You know, I'm not going to make this arrest now because I just saw what happened to my friend. 

 

Felipe [00:46:30] That definitely could be happening. And I don't think our goal is not really to delineate these different types of responses. We're really picking up on an aggregate response across these different types of stories. What I would say is that the first story here where the officers know what arrests don't have any crime reducing benefits and they stop doing them and so crime doesn't go up that one ok there's no crime increase. The second one, maybe you would think there should be a crime increase from that that story. So you think that if the officers are now no longer arresting the big, scary guy who's going to commit a new offense, then you would expect that crime is going to go up. The fact that we don't see a crime increase suggests that maybe it's the first story, or maybe that if we're worried that some policy change induces this second kind of story, it's maybe not that big of a concern for crime increases. 

 

Felipe [00:47:24] So even if you think the second dimension is actually part of the response, it doesn't actually seem like it's causing a crime increase. 

 

Jennifer [00:47:32] Yeah, that's an excellent point. Yeah. And I think my my hunch is the same as what I think yours is that it's mostly these low-level offenses that they're just they're just not bothering arresting people for trespassing anymore, which maybe doesn't matter. And that's useful to know. 

 

Felipe [00:47:47] I will note that, you know, we do find reductions in violent crime arrests and serious property arrests. So these are the type that you wouldn't think that would be the first thing for our society to say. Let's stop making these arrests, right? Probably any kind of policy that's initiated in terms of, say, decriminalizing some kind of offense is not going to decriminalize some kind of violent offense. 

 

Felipe [00:48:10] But we see reductions in arrests for violent offenses. There doesn't seem to be a crime increase. I do think this speaks to the point that even these second types of responses are maybe not damaging for public safety. 

 

Jennifer [00:48:22] Mm hmm. Yeah. And they might still be marginal in some way that is relevant to that conversation. OK, so bigger picture. What do you think of as the policy implications of your results? What should policymakers and practitioners take away from all this? 

 

Felipe [00:48:36] So the first thing that I would say here is that what we're trying to do with this study is kind of delineate the different dimensions of policing and emphasized that police enforcement activity is different from police employment and police presence on the street. And so this is kind of a subtlety that we think is important in the policy discussion here, which is that separate from the number of officers you have in your police department, the number of arrests they make is a very important question for society. So arrests have very big social costs, not just for the people who are arrested, but also for their family members, also for the sense that people have how trustworthy the police are. 

 

Felipe [00:49:25] So there's a lot of research documenting that there's social cost of too much enforcement. And what we think of as the big policy takeaway here is that at least at the margin, you can reduce the amount of enforcement activity for low-level offenses holding fixed the number of officers you're hiring. So we're not saying, you know, fire police officers, we're not saying reduce the number of officers on the street. If you reduce the number of arrests officers are making, you likely won't have a damaging effect on public safety. So that's the big takeaway here is that let's go into the specifics of what dimension? Of policing, we're going to change, and you can actually have police officers on the street corner not making arrests but still deterring crime simply by their presence. 

 

Felipe [00:50:14] So that would be, I think, the big takeaway here from our study. So let me emphasize here again, do I know I've used this word already? We're talking about marginal enforcement. So what we're saying is at current levels of arrest activity, you could reduce arrests for low level offenses by a small amount and not sacrifice public safety. We are definitely not saying stop making arrests as a whole, stop making arrests for entire offense categories. This is definitely a marginal statement here. I think that's important that we don't kind of paint with a broad brush about what lesson we're trying to draw from this study. 

 

Jennifer [00:50:54] Yes, err on the side of greater leniency, which seems to be the punch line of a few different studies coming out these days, which is always nice when there's a big overall message. What other papers have come out that are related to this topic since you all first started working on this study? 

 

Felipe [00:51:09] So I'll mention two papers that I think are fairly related to what we are doing and that have come out recently. So one paper is by Aaron Chalfin, David Becerril and Morgan Williams. This is a paper that came out this year looking at a similar event in New York City. So a few years ago, there was a pullback in enforcement by the New York Police Department for about a month. There was kind of a union dispute between the police department union and the city, and through that the department the officers were kind of disgruntled and they reduced their arrest activity for about a month. And what Aaron, David and Morgan find is that this did not lead to an increase in crime or really any change in crime and no increase or decrease in crime. 

 

Felipe [00:51:58] So we think that this is very related to our study, which is that you're looking at this event. That's not really a policy. It's an officer initiated change in enforcement. Maybe they're using their discretion to decide what are the types of arrests to no longer make, and there doesn't seem to be a change in crime. So I think of that as having a very similar take away to our study. The other study I'll note here is your paper with Amanda Egan and Anna Harvey on misdemeanor prosecution. You can talk about this much better than me, but since I'm the one answering the question, I'll go ahead and describe this paper, which is that you guys look at what is the impact of a prosecutor not charging an arrestee with a misdemeanor charge. So prosecutors get these arrests that police officers make. They're assigned to a prosecutor and they decide whether or not to go through with it and charge the individual. 

 

Felipe [00:52:54] And what you guys find is that individuals who are assigned to a prosecutor that is more lenient so has a greater propensity to kind of drop all the charges. They actually have better long run outcomes. They're less likely to commit an offense after this incident, which is surprising. At least I found it surprising because what it's suggesting is that these misdemeanor prosecutions not only do they not have a crime reducing benefit, they actually seem to have a criminogenic effect. So you get a misdemeanor prosecution or you're charged with a misdemeanor offense and you commit more crimes in the future. So I would say that that's even a stronger, more stark finding than what we find, which is not only do some of these low level sanctions not have a crime reducing benefit, they could even backfire. 

 

Felipe [00:53:44] So that's even stronger evidence for why we might want to pull back some of these enforcement measures at the margin. And, you know, in line with some of this discussion, we're having about discretion here. This is also a study that's using kind of discretionary variation, rather than some stark policy change that changes what is legal or illegal. 

 

Jennifer [00:54:04] Perfect summary. 

 

Felipe [00:54:06] OK, great. 

 

Jennifer [00:54:08] And what's the research frontier? What are the next big questions in this area that you and others will be thinking about in the years ahead? 

 

Felipe [00:54:15] So I think that there are a couple of different ways that this research area is going to go moving forward. So one of those is going to be getting into the specifics of which types of arrests are most important for deterrence. So to reiterate, we know a lot about the effect of police employment. We know a lot about the effect of police presence. We don't really know as much about these specific types of enforcement activity. And these studies, like my study, Aaron and coauthor studies, these studies are looking at broader changes in arrest activity. I think it's going to be important to figure out what's the treatment effect of these various dimensions of enforcement and various types of arrests. Like I said, I doubt somebody is going to take the study and say, Well, we see violent arrests are going down. Let's decriminalize some violent offenses. So getting into the specifics of are the effects different across these different types of categories, I think is going to be an important future question to answer. On the policing side I will go back to what you said before, which is that there's really not a lot of research going into the specifics of what police officers do day to day. 

 

Felipe [00:55:23] A lot of research on policing up to this point has been using department level analysis. That's largely the realm in which we operate in this study is using department level analysis, and I think in the future there's going to be a lot more research looking at individual police officers looking at the specific choices that they make throughout the day. We speak to that a little bit in the sense that we are studying how police officers respond to these events, but I think that's going to be a growing area, how two police officers spend their day. What are the determinants of how they choose, what enforcement activities they engage in? 

 

Felipe [00:56:00] On that note, I will cite one working paper of mine with Aaron Chalfin, where we try to do a lot of this looking at how police choose their arrest activity throughout their work shift. And so this is a study called the Professional Motivations of Police Officers, where we're looking at the potential impact of labor market considerations or labor considerations of officers and how they choose their arrest activity. And so we look at the hourly arrests activity of officers and how is that a function of whether or not they're going to get overtime for their arrests, whether or not they have an off duty job at the end of their day that they're planning on going to? And how do they balance that with more, let's say, altruistic considerations of wanting to do their job properly? So I think are going to be a lot more studies on that type of question as you get better data on policing through their own initiative and through changes in laws, the data availability is increasing on policing and crime data. So I think that's going to be another big change here. 

 

Felipe [00:57:03] Yes, I would say those are the two big areas is specifics of what types of arrests are most important for deterrence. And second, getting under the hood of what police are doing throughout their workday. And what are the various determinants of their police activity? And as an offshoot of that, I would say thinking about what kind of policies do we want to use to improve police behavior? Are there personnel interventions that might make sense for improving police behavior? 

 

Jennifer [00:57:30] Lots of exciting work going on in this space thanks to you and your colleagues, and I'm excited about all the new data, too, so I will definitely be watching for more. My guest today has been Felipe Gonzalez from UCLA. Felipe, thanks so much for talking with me. 

 

Felipe [00:57:44] Thank you. 

 

Jennifer [00:57:51] 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 and other contributors. Probable causation is produced by Doleac Initiatives, a F01(C3) nonprofit, so all contributions are tax deductible. If you enjoy the podcast, please consider supporting us via Patreon or with a one time donation on our website. Please also consider leaving us a rating and review on Apple Podcasts. This helps others find the show, which we very much appreciate. Our sound engineer is John Keur with production assistance from Nefertari Elshiekh. 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 DoleacDifference-in-Differences, Policing