Episode 55: Morgan Williams, Jr.

 
Ep55_Williams_photo.jpeg

Morgan Williams, Jr.

Morgan Williams, Jr. is an Assistant Professor of Economics at Barnard College.

Date: August 3, 2021

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

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Prof. Williams's work on the race-specific effects of policing:

"Police Force Size and Civilian Race" by Aaron Chalfin, Benjamin Hansen, Emily Weisburst, and Morgan Williams, Jr.

 

Transcript of this episode:

 

Jennifer [00:00:08] Hello and welcome to Probable Causation, a show about law, economics, and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an Economics Professor and the Director of the Justice Tech Lab.

 

Jennifer [00:00:18] My guest this week is Morgan Williams Jr. Morgan is an Assistant Professor of Economics at Barnard College. Morgan, welcome to the show.

 

Morgan [00:00:25] Thank you for having me, Jen.

 

Jennifer [00:00:26]  Today we're going to talk about your research on how hiring more police officers affects local crime outcomes. But before we get into that, could you tell us about your research expertise and how you became interested in this topic?

 

Morgan [00:00:38] Sure thing. So thank you again for having me. And I'm an economist. And so one thing I like to kind of say to people when I describe what it is that I work on is the fact that well, I consider myself an economist and scholar of race first and pretty much everything else second. And so at the moment, I kind of see much of my research agenda kind of thinking about the nature of racial disparities through the prism of criminal justice. And so we know that violence and crime can have very serious and kind of detrimental societal costs that could look very different for one community the next. No shortage of research has shown that economic development is deeply affected by this: health, cognitive development. The list goes on. And so in many ways, I think that many of us who are students of the economics of crime in some way always have to go back to that 1968 Bekker model, which kind of gives us some insight on the nature in which we can kind of have a bit of a say on the prevalence of crime in many of our communities.

 

Morgan [00:01:37] And so in reading through some of that work, I always kind of said to myself, well, OK, well, you know, we certainly can affect certain parameters like the probability of apprehension and also fines and do those things that can allow for us to enjoy certain crime rates. But is that true across various types of criminal behavior? Is it equally true for, say, robbery as it is for homicide, to the extent that police and the courts matter and are major factors in determining some of the outcomes that we're interested in? Well, are they efficient? Are there any collateral problems? Can they have additional criminogenic effects that we're not considering? Are beliefs important? Stereotypes? These are things that all kind of influence the types of criminal behavior that we see and also what type of criminal behavior we focus on. So, as a result, you kind of step away from some of that work and you say, OK, well, these are certain ways to think about it. There might be other ways to think about it. In fact, there might even be other policy interventions that we might want to consider. So you might want to think about things like gun control. So you might want to think about other types of interventions that can be pretty important in determining the nature of, say, like homicide or some other type of criminal behavior in ways in which maybe we didn't foresee.

 

Jennifer [00:02:46] So your paper is titled, "Police Force Size and Civilian Race." It's coauthored with Aaron Chalfin, Ben Hansen and Emily Weisburst. And last I checked, it's been conditionally accepted at AER Insights. Is that right?

 

Morgan [00:02:58] That is correct.

 

Jennifer [00:02:59] Excellent. So we are obviously in the midst of a serious national conversation about the role of police in communities, and this paper contributes directly to that policy conversation. So what are the mechanisms through which we might expect that expanding or shrinking police forces might affect crime outcomes?

 

Morgan [00:03:17] It's a great question. And luckily for us, we were able to lean on a very robust and growing literature that seems to be suggesting that increasing the police force size can have very important ramifications for criminal outcomes. And so you can obviously go back to see Levitt's work in 1997 that kind of linked local and state electoral cycles, police staffing levels and which found significant reductions in violent crime. Justin McCrary then replied to that work and said maybe the inference isn't quite right. But Steve's replies seem to be very convincing when he uses firefighters as an instrument, as opposed to leaning on the variation that might be attributable to, say, to local and state electoral cycles. And so that's kind of like one paper, right?

 

Morgan [00:03:59] But then we also have a host of literature that's taking a look at the COPS grant, so the Community Oriented Policing Services Office's grant, which you'd be surprised how many people just don't realize that these COPS grants, sort of the COPS office, was established under the 1994 Crime Bill. And so with that bill, the goal was pretty much as it seems, was to get more- pretty much more police on the streets in cities throughout the country. Essentially, what they do is they issue out block grants. I believe about more than half of them are pretty much hiring grants—being able to beef up on loan officers and other personnel—but also the ability to invest in nonhiring aspects of policing as well, so technology, other kind of targeted crime initiatives, things that are all kind of contributing to the nature of public safety.

 

Morgan [00:04:44] And so with those grants, there's been a good bit of work starting back with Evans and Owens in 2007, which again, kind of exploited the nature of these grants or the size of these grants in order to be able to say a little bit more about policing staffing levels conditional on fixed effects. And the ability to account for some of those other things that might be different across agencies, but also might be some common shocks that maybe existing during this time period. So they also ended up finding that there seems to be a reduction in burglaries, auto theft, robbery, aggravated assault and probably more measured evidence on murder as well. So that was kind of a very good sign that this literature seems to be going in the right direction.

 

Morgan [00:05:25] Now, there's been subsequent work that's also kind of exploited these COPS grants, either in very similar ways or maybe in slightly different quasi experimental ways. And so, Emily Weisburst, my coauthor, has also kind of used a very similar type of methodology. And in doing so, she was able to kind of capture other aspects of the hiring process and also kind of look at a later period in order to be able to kind of come to a similar conclusion about reductions in violent crime and property crime. I think that she had- she found a reduction in violent crime of about 13 percent and a reduction in property crime of about seven percent.

 

Morgan [00:05:58] Now, one thing that I probably should back up and say in kind of discussing the COPS grants, is that they pretty much were fairly prevalent during the Clinton administration, heading to the late '90s. Pretty much anybody that wanted to apply for a grant did receive some funding. But there were concerns starting around the Bush administration that much of the funding that was coming from COPS was actually supplanting the existing funding that would have been used to hire more cops anyway. And so around the Bush administration, we do see the COPS grants tending to dry up a little bit. But heading back into the Obama administration and after the Great Recession led to some very important fiscal pressures on many law enforcement agencies throughout the country. The COPS grants ramped back up. They were a bit more competitive. There was probably a bit more kind of leaning towards making sure that police departments weren't hemorrhaging officers more than anything else. So that's kind of like a kind of quick recap on maybe some of the policy elements of COPS.

 

Morgan [00:06:52] But then we've also seen other quasi experimental approaches that I mentioned before. So Steve Mello had a paper and his colleagues also had a paper that kind of used a regression discontinuity design, kind of leveraging the fact that rating scores are used in this latter period in terms of issuing the COPS grants. And that also kind of allows for you to kind of have some credible estimates that are very consistent with many of the papers that I've seen- I've kind of mentioned before.

 

Morgan [00:07:17] I'll mention one more paper, but I'll kind of reserve maybe some of the discussion about the mechanics behind it. But the reason why we're kind of doing this is because we're merely interested in the fact that well, perhaps there might be other reasons that police staffing levels might change. Right. And so it could be simultaneity wise, could it be the police kind of forcing crime waves or are they driven by crime waves? Is there a omitted variable problem? Well, there's another paper by my other coauthor, Aaron Chalfin, and Justin McCrary that kind of makes a different argument in which they say, well, wait a minute. Well, maybe there might actually be a measurement error issue. Maybe that measurement error issue when it comes down to police employment levels might be more important, right, in terms of the estimates that we're able to obtain. And so they argue that in addition to the presence of measurement error, there are also very important institutional constraints that exist in the manner in which police departments are able to kind of ramp up hiring. It's not as if they could just say, you know what I want 100 officers on the street tomorrow, snap their fingers and they're there. These things take time. It takes time to bring people into the academy. It takes time to train them. And it probably takes probably what I believe it was somewhere between six months to a year or longer in order to kind of get those cops on the street. And so they also take a similar approach in which they say, you know, let's kind of just look at two different measures of police employment and they actually use a very similar instrumental variable strategy and find comparable results. But also find is that- it seems as if many of the cities within their sample also suffer from under policing according to their social ---.

 

Morgan [00:08:48] And so we have this large literature that we were able to kind of lean on in order to be able to ask a very similar question about when it came down to race. Because one thing in which when you talk to people, many people kind of say, well, why does policing staff staffing matter, at least in the name of public safety? Why does it matter in different ways for, say, Black and white communities? And so in my head, I could think of a number of different reasons why that might be the case. It could be the case, in terms of the benefit side, that perhaps there just might be more opportunities to solve crime. And so it could just be the fact that, well, within the United States, Black Americans make up about 13 percent of the population, but over 50 percent of the homicide. So police are going to have more opportunities to kind of participate in crime reduction, perhaps, in Black communities.

 

Morgan [00:09:34] There's also kind of disproportionate levels of street bias that might exist across one community to the next. And it certainly is the case when you look at it across racial groups. And so with that difference in visible street bias, many of those markets, they kind of operate by different rules. If you have a dispute within a market, say within a drug deal or something like that, it's not as if you can go to court and say, hey, officer, he stole my drugs. It's one of those things in which there obviously are other ways in which these types of disputes are kind of handled and ways that kind of escape our traditional market explanations. And so one of those explanations often been given around the crack cocaine epidemic and elsewhere has generally involved violence. Right. And so to the extent that you have street bias that are- have these very kind of robust and prevalent markets that are accompanied by violence again and to the extent that they are kind of geographically concentrated within Black communities, well, then that also kind of gives you another reason why it could differ for Black Americans versus white Americans.

 

Morgan [00:10:31] And there's also just the idea that perhaps there just could be some additional cost that we're not thinking about as well. We know certainly for sure that there's a good bit of work on hotspot policing and within that hotspot policing, we're starting to see that again you can kind of target areas and space in successful ways in order to reduce criminal behavior. But when you do that, there's also kind of changes in community perceptions over police legitimacy. So with those changes in police legitimacy, that's something that could have some larger kind of effects on the ability of law enforcement to kind of affect crime. And so there's no shortage of reasons why we kind of felt as if policing staffing levels could have important consequences for racial disparities in things like homicide and arrest patterns. So we were able to kind of leverage this larger literature in order to be able to kind of say something meaningful about it in an empirical way.

 

Jennifer [00:11:23] Yeah, so let's talk more about why we didn't know more about these effects by race. It seems like such an important question, an important issue given the policy context. It's certainly something that, you know, has been on people's minds for a while. But you're right that this this huge literature looking at the effect of police on crime has really focused on measuring the overall effect, the average effect across everybody in all communities. So why don't we know more than we do? Were the challenges primarily data related, or do you think it's something else?

 

Morgan [00:11:53] Well, to be honest with you, I kind of think it might be a combination of- one, it might just be a blind spot of the literature. And so there's no shortage of work out there that kind of takes a look at these more aggregate patterns. But as somebody as I mentioned at the top of the podcast that focuses primarily on race, I tend to see that this exists in other areas of the literature as well. So let me give an example. So with gun control policy, we've had no shortage of papers generally that kind of revolve around natural experiments, around changes in gun laws that tend to look at the impact of say background checks or concealed carry law and their impact on homicide, victimization, and suicide. Almost all of those papers—obviously, we have some very kind of established scholars that kind of tackle those papers and work through those papers—more of those papers kind of focus on these things in an aggregate way.

 

Morgan [00:12:42] But for me, I tend to look at some of the descriptors. I tend to kind of hearken back to that Becker model. And I just say to myself, well, when we take a look at the descriptive evidence, there is no shortage of evidence- doesn't suggest that much of the patterns that we see in robberies, much of the patterns that we see in homicide, have some type of racial element to it. We live in a largely segregated society. And so we would expect for the social interactions that might generate, say, homicide and homicidal interactions, would tend to be a bit more personal, or robberies which tend to be a little bit more impersonal, could lead to the complexion of those interactions being a bit different. And therefore, maybe we should be thinking about race. The same thing is true for gun control policy and some of the other things I've mentioned. And so why exactly we don't necessarily have that is I think is probably a question I would toss back to the larger literature. But I do feel as if the kind of underlying social interactions and how they're kind of influencing the types of crime waves that we see over space and time is something that we should all be thinking about when conducting this work.

 

Jennifer [00:13:41] Yeah, I think you're completely right that this is this is one reason we need a diverse set of people at the table with all the questions that they might, you know, that they think are important, asking the research questions and doing the research. And it's probably something that just hadn't been front of mind for the type of person who was an economist usually in the past. I do think that like as I'm thinking through the data that exists, we're always a bit constrained with the data. Clearly, identification's not as much of a challenge here because there is this large literature that has figured this out, right. And you're going- we're going to talk more about the strategies you're using later, and so that- you know, there are kind of paths forward there. On the data, I think people probably have worried about the extent to which race is sort of selectively reported, especially for crimes that aren't always reported to police. So one way you're going to handle that is by focusing on homicides. So that brings us to data. So why don't you talk about the data you're using for this?

 

Morgan [00:14:41] So it's another kind of very good point. Right. That I should probably highlight in- or as a caveat to kind of what I just said. So when it comes down to understanding the importance of race and policing and all of these other things, a big part of that has to do with like, well, one, what is the nature of your empirical design? So if you're going to focus on thousands of agencies, I imagine if you're going to focus on even something as well reported as homicide, you might not necessarily have the complete data necessary in order to be able to answer that exact question across racial lines. That certainly is a point well taken.

 

Morgan [00:15:13] I think it also kind of depends on the question that you're asking as well with respect to crime. And so you made the very astute point that homicide is something that we have a better kind of grapple, right. And so we generally have two different sources of homicide data: one coming from reports from law enforcement agencies themselves in which they are reporting much of this information in the uniform crime reports. And also you can find it within the supplementary homicide reports, which are incident level data. But much of that is kind of reported, self-reported or reported to the police or reported by the police. And so as a result, there might be some concerns about the nature of the coding of race in some data, depending on how large the sample you work with.

 

Morgan [00:15:57] Now, one of the good things that we can lean on is the fact that we also have death certificate data. So we also know from your local Health Department or your State Health Department and subsequently the CDC and other agencies that we can get a sense of trends in homicide rates according to these death certificates. And as a result, we can kind of look at the series over time—and some of the work that has looked at these series over time tends to suggest that these are pretty closely tracking one another, these two measures. And so as a result, we tend to focus on things like homicide because of the fact that it's better reported in terms of on the victimization front.

 

Morgan [00:16:33] When it comes down to other crimes, it really just depends on kind of what it is that you're going to focus on. So when it comes down to, say, motor vehicle theft, because of insurance policies, people tend to report motor vehicle theft more accurately than maybe some other crimes or things that we don't necessarily get a reporting on all the time, like assault. So we do focus on homicide victimization exclusively. And we do use the data sources that I kind of mentioned before, right, the FBI supplementary homicide reports. We focus exclusively on a sample of about 242 cities from the I would say about the early 80s until more recent years. And so in doing so, we're able to construct homicide victimization rates at the city level per hundred thousand.

 

Morgan [00:17:11] What we also use, and as I kind of alluded to before, is that we need measures of police employment. And so in order to get a sense of police employment, we're going to pull from two different sources. One source that we pull from is going to be the US Census Annual Survey of Government. And we're also going to pull from the FBI Law Enforcement Officers Killed and Assaulted or the LEOKA data. And so these data provide point in time measures of the full time sworn officers, and then we measure that at the city level. And so in doing so, we have two complementary measures of police employment that we can kind of use both as a check, but also as some of the- as part of an empirical approach to kind of credibly estimate the impact of police staffing levels on many of the outcomes that we're interested in. So we're going to include some of those data. We're also going to include data from the COPS program. And so with the COPS grants, we're going to have information on the size of the grants. We're going to have information on other aspects of the application process. So whether it was hiring, non-hiring, those types of things.

 

Morgan [00:18:08] And finally, we're going to include a bunch of census data and also we're going to include enforcement activity data from the FBI Uniform Crime Report. So within these data, backing up and thinking about the enforcement- about the arrest data, and so the arrest that we're pulling from the Uniform Crime Reports, we're going to focus on pretty much two different levels of arrest that exist. So we're going to focus on index arrests, which kind of refer to arrests for index crimes or very serious crime that the FBI tracks very closely. So robbery and aggravated assault, murder, etc. We're also going to take a look at what we call quality of life arrests. And so these quality of life arrests consist of low level misdemeanor offenses that might consist of disorderly conduct or drug possession, generally offenses that police have a great deal of discretion on- over. And we can kind of view as perhaps being victimless in some ways—not to say that people desire these types of offenses within their neighborhood, but we are going to have those two measures of arrests that we're going to look at as kind of like primary outputs.

 

Morgan [00:19:09] And finally, I should also add that we're going to include measures of municipal expenditures. And so with municipal expenditures, what we're going to be able to do in controlling for overall expenditures within our models, we're going to be able to provide estimates of the impact of a marginal officer on homicide on these low level quality of life arrests, on these index arrests. But our measures are going to actually reflect the historical opportunity cost of investing in police manpower versus something else. So we actually see this as a kind of critical aspect of the analysis because it allows for us to contribute to a very real kind of debate that's taking place with respect to defund. Right. So a big kind of concern with defund is the nature of okay so we reallocate X amount of the policing budget to some other agency. We would be in a better position or perhaps we hope to be in a better position. Now, we can answer that question exactly, but we can speak to those historical trade offs that I mentioned before. So that's pretty much covers the nature of the data, all the data for the most part are pretty much available publicly.

 

Jennifer [00:20:08] Yeah, so just to say that last piece back to you in slightly different words, you're going to be looking at the effect of spending a marginal dollar or hiring a marginal police officer, holding constant what the local budget is.

 

Morgan [00:20:22] That is correct.

 

Jennifer [00:20:23] Great. OK, and then can you say a little bit more about what quality of life offenses are? Give some examples.

 

Morgan [00:20:29] Yeah, sure thing. So I gave a few examples- so disorderly conduct is one. Liquor law violations are- kind of serve as another type of quality of life offense. Curfew, loitering, suspicious persons. These are things that for which it might be a bit more difficult to identify a given victim for and for which the police can kind of make- use their own discretion about whether or not to make an arrest for that offense. And so one thing that we kind of like about being able to look at quality of life arrests is because, contrary to much of the earlier literature that I mentioned before, we were able to kind of add a new element or a new wrinkle to kind of understanding the nature of these quality of life arrests. Because on the one hand, perhaps these arrests might lead to a fine, maybe even a jail sentence or neither. On the other hand, it can have an important criminogenic effects. And so by continuously going to court for these smaller offenses, perhaps it ends up entangling other parts of one's criminal history. And as a result, maybe it leads to incarceration or it could be something that could contribute to a larger aspect of public safety. So we want to be able to account for this, but we want to have some type of conversation about what exactly are the police doing when you do increase police force size. And certainly we want them to be making arrests for these more serious crimes like murder, rape, robbery. But we also want to get a sense of how they treat these discretionary offenses, because I think it could open the door for a larger conversation on what the police are and are not doing, how productive they are, and increasing public safety, at least at this measure.

 

Jennifer [00:21:59] Awesome. And so remind us again the main outcome measures then that you're going to be focusing on here.

 

Morgan [00:22:05] Yeah. So the main outcomes that we're going to focus on is we're going to focus on homicide victimization and we're going to exclusively focus on non Hispanic Black and non Hispanic white homicide victims. And the reason for that is that within the data themselves, as we kind of discussed before, there can be some quality issues. Those issues tend to appear more for the coding of Hispanics within the supplementary homicide reports. And as a result, we decided to say, you know what, we're going to just focus on these two groups because they tend to be better reported. And over time, you do see changes in how Hispanics are kind of coded within the data in ways that might not necessarily be consistent with other kind of different verification information that we might have. So we tend to focus on that.

 

Morgan [00:22:45] Now with the arrest outcomes, we are actually going to, again, focus on index arrests, those more serious crimes that I mentioned before, in addition to the quality of life arrests. Now, with the arrest data, we cannot focus on ethnicity, we cannot account for Hispanic ethnicity, to be more specific. But in thinking about this, one thing that we do know from the larger social science literature is that for Black arrestees or for other measures that are kind of similar to this situation, we tend to not see much change in how race is kind of reported among Black victims or Black arrestees. Now, when it comes down for the white arrestees, then it differs a little bit. It can differ by region, depending on how many- what the Hispanic representation is within that given region. And so as a result, the white arrest rates that we end up constructing could have some Hispanics in them, they could not. But I think if anything, that's to our favor in how we're kind of going to be able to kind of speak to where these effects on arrest are going.

 

Jennifer [00:23:41] Yeah, so if we think of the white arrest rates—and probably even to some extent the white homicide rates—as being sort of some mix of of white and Hispanic, that should probably bias you toward finding results that are closer to the Black estimates if we think of Hispanic residents as being sort of somewhere in the middle in terms of how they interact with police officers.

 

Morgan [00:24:00] That is exactly correct. Yes.

 

Jennifer [00:24:01] All right. And so and you also mentioned you're using data from 242 large U.S. cities. Tell us a little bit about your sample. What do those cities look like and how do they compare with other places?

 

Morgan [00:24:12] So, for our sample and focusing on 242 cities, one thing I'll note is that our sample, when you look at some of the earlier cop literature, they kind of use a much larger sample of cities. Right. So they're including a lot more smaller cities within their sample. For us, we're going to include- focus on these large cities because we're able to kind of measure these things, all these kind of variables of interest to us in in credible ways. But then also in doing so, we feel as if we can kind of speak to this race issue a bit more closely, as we kind of discussed before. And so within our sample of 242 cities, it looks- it's obviously fairly urban. I believe the number of homicide victims, obviously there's a large disparity there. I think it was about 139 I think Black victims on average within our sample—we weight by population in 1980—compared to 64 for whites. We also see a difference in clearance rates. Clearance rates are a bit higher for whites. And also, these cities are generally maybe have a larger Black representation, so I think maybe about at least on average, 24 percent are Black compared to 48 percent white. And so and looking at these numbers, we can kind of say to ourselves, we're speaking to a very large urban environment throughout the country.

 

Jennifer [00:25:24] Alright. So walk us through your empirical approach. How do you measure the causal effects of police hiring on these various crime outcomes you're going to be looking at?

 

Morgan [00:25:33] Yeah I mean, that's great. So, I mean, one way in which we do so is kind of it's pretty much leaning on the literature that I mentioned before, respect to the COPS grants. And so what we're going to do is that we want to be able to kind of have- in order- if we think simultaneity bias is a problem or if there's an omitted variable, we want to be able to have some sort of variation in police staffing levels from one year to the next across agencies that can be explained by being awarded a COPS grant. Right. And so in this way, what we end up doing is that we kind of employ an instrumental variable strategy similar to the rest of the literature. What we end up doing is that we say to ourselves, OK, can we understand the impact of hiring one more officer due to being awarded an officer through the COPS grant? And can we examine that impact on, say, homicide or arrest? In doing so, what we end up also doing is that we also had a few controls for much of the application process. And so as a result, by adding controls for the application process, whether or not they apply- the agency in question applied for some type of technology inspired grant or something of that nature, we're able to say, OK, well, maybe there are certain characteristics that are associated with those interests that allow for us to control for things like quality. We also are able to control- you obviously got to have some fixed effects in there and that's going to allow for us to condition on those fixed effects that's going to allow for us to kind of be able to kind of exploit this COPS grant's plausibly exogenous variation.

 

Morgan [00:26:58] Now, what we also do—and this is kind of another study by Aaron and his coauthor Justin that I mentioned before—basically they say to themselves, OK, well we might not have an experiment, but what we can do is say, let's take the two policing employment measures that I mentioned, the one that is attributable to the ASG and the other one that's coming from the LEOKA data. And so if we believe that there might be an issue of measurement error—and I do kind of quote the paper and go through an example in New York where it was just like this huge change from one year to the next that it doesn't necessarily make a lot of sense—is say alright if classical measurement is indeed an issue, then what we can do is that we can use one of these variables as an instrument for the other in order to be able to explain changes in police staffing levels within our sample and their impact on homicide and arrest patterns. And so in doing that, we're able to kind of provide complimentary identification strategies. And if they're close- if the estimates are close—and we show in the paper that they are—then that allows for us to have a bit more faith in the manner in which we believe policing staffing levels are working. And so we also kind of see this as an important contribution as well.

 

Jennifer [00:28:03] Yeah, it is really interesting that there are these two approaches and there- for so long, especially in the economics literature on policing. And part of the reason there's such a large literature is there has been such a focus on finding a good instrument or finding a good natural experiment to identify the causal effect of police on crime. Because there were all these early papers, right, that showed if you just look at the correlation, it actually looks like police cause more crime because police departments tend to hire more police officers when crime is going up. And so you wind up finding this positive correlation. And folks looked at that and said, well, that could just be because it goes the other way. There's a simultaneity bias you're talking about. And so all this energy went into finding these interesting instruments and stuff. All the, you know, let's use the COPS grants. You have to get funding first or whatever else. And then Aaron and Justin came along and it was like, well, it's mostly measurement error. So we can just use an IV to correct for the measurement error, which is also really cool and fancy, but I do think it's just fascinating that we basically get the same result. So, yeah, so you're going to be using both of those approaches. And I think if I recall from what you say in the paper, you're basically going to find the same results.

 

Morgan [00:29:10] Yeah, for the most part. Now one thing I will say is that because the COPS came about in the mid 90s, we have to kind of shorten the sample period that we work with. But when you restrict the sample period, depending on the outcome, they tend to be very, very similar, if not statistically similar, and tend to at the very least provide very consistent qualitative stories about the results. And so we think that's some good guidance for how to conceptualize or maybe having resolving some of these larger arguments that you just mentioned before about the nature of how credible are some of these identification strategies that the literature has completed.

 

Jennifer [00:29:42] Alright. So let's talk about the results. What do you find is the effect of police hiring on homicide rates for white and Black residents?

 

Morgan [00:29:51] So basically what we end up finding, and kind of focusing on homicide, we find that the marginal officer abates about 0.06 to about 0.1 homicides, depending on which model we use. Another way of thinking about that is that in order to save one life, it requires the hiring about 10 to 17 more officers on average, and so in that way, one thing that's interesting is that while you look at the estimates across Black victims and white victims, statistically they're pretty much the same. Right. And so the marginal officer seems to save just as many Black lives as white lives. But one thing that we kind of noted already, based on our standpoint also true of the country just in general, is that Black Americans make up a small proportion of the population. So why don't we also kind of look at these results in per capita terms as well? Right. And so we want to be able to kind of benchmark our estimates to the population to kind of address the issue that I mentioned before.

 

Morgan [00:30:43] And in doing so, we actually notice that the gains are about twice as large— gains in homicide victimization that is, I mean, in terms of the reduction—are about twice as large for Black victims on a per capita basis. And so one thing that's kind of interesting to note here is that while I think within economics, we tend to- I think we've moreso kind of gravitated towards the idea that police do have an important role in reducing violent crime. But when you step outside of economics, this is actually been a point in which many people seem to not necessarily be cognizant of. And so one thing that kind of takes that idea a step further is the fact that not only do police reduce violent crime, but they also reduce the disparities in violent crime that are just so crucial to life expectancy and other kind of societal considerations that we have when it comes to homicide. So I do think that's a pretty important finding both within the economics of crime literature and at large.

 

Jennifer [00:31:35] Yeah, I totally agree. And those effect sizes are per year. Is that the way to think about it?

 

Morgan [00:31:41] Yes.

 

Jennifer [00:31:41] OK. Yeah I mean, those effects are just huge. I mean, I know the literature and it's still- every time I see an estimate like that, it's like that's a lot. I mean, the idea that, you know, every officer you hire, you'd have to hire 10 to 17 officers to save one life. The other way you think about it, I guess, is like if you hire one officer over the course of like a 10 year span of his being on the force, he's going to save a life for sure. And that is just fascinating.

 

Morgan [00:32:04] It is kind of steps outside of what people's original considerations- like a lot of times when we're having this kind of larger kind of police reform debate, they'll say, well, police actually spend most of their time in their car like on- waiting for the radio to go off or something-.

 

Jennifer [00:32:19] Buying donuts.

 

Morgan [00:32:19] Yeah, eaint donuts. I will kind of comment on a bit later, but one thing that two of my collaborators, Brenden O'Flaherty at Columbia and Rajiv Sethi at Barnard, said was that you could even put cops out on the street and they could just have a donut eating effect. They're just eating donuts and they can still solve crimes. So I think that is kind of important to know that they do have an important role in addressing homicide victimization. And that is kind of a very important consideration I'm not sure is discussed enough when talking about the role of the police or what it should be.

 

Jennifer [00:32:49] Yeah, it makes all the conversations much more complicated if you're not happy with the way police are behaving. OK, next set of results. So you also consider the effect of police hiring on arrests for serious crimes, these index crimes you mentioned before. So what do you find there?

 

Morgan [00:33:03] When we look at, you know again the same measure, what we end up finding for index crimes is actually arrests for serious index crimes fall with larger police forces, which many people kind of scratch your head over to kind of give a sense of how I kind of interpret this in a second. But, you know, basically the marginal officer makes about 1-2 fewer index arrests on average. And one thing that kind of makes it interesting, in thinking about arrests, arrests are going to be a function of policing strategies. They're also gonna be a function of crime that exists. Right. Or what- perceived crime. And so we need something to be able to kind of benchmark this to. And so what we do is that we also say, well, we don't have a race specific measure for this, but let's also take a look at the index crimes themselves, not the index of arrests but the index crimes themselves. And so as a result, we end up seeing that the marginal officer actually abates about 18 to 24 index crimes. And so you have an average decline in arrests for index offenses. You have an average decline in index crimes that's associated with expanding the police force size.

 

Morgan [00:34:03] And so one way to kind of look at this is that it doesn't seem as if police are kind of arresting their way out some of these public safety gains that seem to be taking place throughout our sample period. What they seem to be doing, though, is that perhaps there's a deterrence effect that might be associated with increasing the police force size. And so when we disaggregate our results across a number of different types of index funds, what you end up seeing is that many of the index arrests that are kind of being reduced are things that you would think or at least would make logical sense, kind of associated with police having a visible and physical presence in greater numbers. And so reduction of property related offenses, theft, and things of that nature, motor vehicle theft, these are all things that we're in- were being reduced the most when it comes down to increasing police force size according to our estimates.

 

Morgan [00:34:54] One thing that I think is also interesting, again, is the mere fact that while- we're a race paper. Right. And so in thinking about the racial disparities, one thing that's interesting is that the reduction in arrests is about 4-6 times greater for Black Americans. And so if you have this decline in crime and you also have this decline in racial disparities, in serious index arrests, perhaps because of the fact that many of these arrests are also accompanied by these lengthy prison sentences—right, because these are very serious crimes—perhaps we're also kind of reigning in the reach of the incarceral state. And as a result, perhaps we're having some by increasing police forces we're also kind of reducing incarceration disparities at the same time. So that's something that we think is kind of an interesting kind of take away from our results on these index arrests and index crimes themselves and the extent to which police force size can also kind of reduce perhaps the incarceral reach of the justice system.

 

Jennifer [00:35:48] Yeah, I agree that is super interesting. And so now to get to the flip side of that possible benefit, you consider the effect of police hiring on arrests for these lower level quality of life offenses, so tell us about that result.

 

Morgan [00:36:00] Yeah. So one thing that we obviously also do is look at these low level kind of misdemeanor offenses, or quality of life offenses. And what we do know is the marginal officer seems to make about 7-22 additional quality of life arrests, depending on which model we employ. And so one thing that I think I will probably note thinking about those quality of life arrests is the fact that well most of the disparities that kind of exist in these additional quality of life arrests seem to be taking place among liquor law violations and also drug possession arrests. And in thinking about the racial aspect of it, I think the effect for Blacks is—at least for liquor law violation and drug possession arrests—was about three times larger than that of whites. And so obviously, you know, there are a number of different kind of considerations, I think at least three that I can think of off the top of my head.

 

Morgan [00:36:46] The first one is that, well, quality of the life arrests, they could be important. They could contribute to the larger kind of breach of the criminal justice system by bringing people in for court appearances and also perhaps tying them up with a former criminal history. In that way, these things could end up being something that is considered a very important cost and consideration. I think the second thing that we want to kind of keep in mind, though, is that because this paper kind of works at a more aggregate level, we don't really kind of speak to preferences. Right. So we're not necessarily putting any structure on preferences. And as a result, it could be the case that many of the Black communities within our sample are just demanding that police do something about liquor laws, violations, and drug possession. And so it could be that these are things that it could be desirable. Right. And so we don't want to rule that out.

 

Morgan [00:37:32] And one thing that I also will kind of say is that on the flip side, we're not able to, at least within this design, ascertain whether or not these quality of life arrests have any deep and meaningful impact on public safety. Now, what I will say is that just based on the literature and a number of studies or policy case studies that we use, whether it be the Floyd case and the ending of Stop, Question and Frisk as we know it, there's just so many different instances in which we kind of see that these quality of life arrests or these low level arrests, might not necessarily have a great deal of productivity behind them. And so we don't want to necessarily say whether or not our estimates speak to the usefulness of these quality of life arrests, but we do kind of want to make all three of these considerations known.

 

Jennifer [00:38:15] So lots of considerations, lots of potential caveats. But as you sort of look at this set of results, what do you think we're learning here about how these effects on arrests or what what they're telling us about police behavior, what are we learning about how they're having such a big effect on homicides? What's your takeaway?

 

Morgan [00:38:35] So I guess my takeaway is more so that one thing that we did not find any evidence for, we didn't find that increasing police force size in either of our models seem to lead to any meaningful impact on clearance rates. And by the clearance rates, obviously, we mean the ability of the police to find- arrest someone that might be responsible for an offense offense or maybe convict them, depending on who you speak to about their definition of a clearance rate. But for us, the ability to kind of clear cases within the SHR data is pretty much our measure, but we don't find any evidence for that. So it doesn't seem as if the police are any more productive in the solving of crimes. And this is something that I think is very consistent with a favorite book of mine, Ghetto Side, by Jill Leovy, that kind of speaks to these very difficult homicides, often in Black neighborhoods, I believe she was focused on Watts, but you have to really kind of get boots on the ground immediately. And there's a great deal of kind of human capital that kind of is associated with being able to kind of get through these homicides and clearing them and also the priorities of departments.

 

Morgan [00:39:37] And so what we can say is it seemed as if the police force size is working through a very kind of sensible mechanism through many of the arrests that we see. Right. So when we look at the arrests, these are things that if a police were on the street, they would be able to stop or perhaps deter. And so in that way, we feel as if perhaps this isn't necessarily police becoming more productive. We can't necessarily say this precisely, but we don't see at least the moral of it as the police are becoming more productive. If anything, they're kind of deterring future crime through their presence. And so if we can achieve homicide reduction both in overall terms, but also in terms of racial disparities, we can also rein in the kind of the incarceral reach of the state. And if we can do this primarily to deterrence, I actually think that this is something that could end up being a very important policy lever. And one that many municipalities already kind of lean on when they decide, hey, I want to do something about crime within our neighborhoods. And kind of creating a holistic picture of this, we do feel as if there are some important takeaways from this research that could be very vital to kind of understanding exactly how we should define the police's role moving forward.

 

Jennifer [00:40:45] And so what should policymakers do with this information? What are the policy implications here? You're finding some deterrent effect of the police that maybe they're sitting in their cars eating donuts, maybe they're walking around arresting people for drinking on the sidewalk. Through a bunch of stuff, they're just like being there seems to reduce serious crime, especially serious violent crimes. But there are all of these concerns about communities being overpoliced and people being harassed and being arrested for stuff that maybe we shouldn't care about. So policymakers that are thinking about this and have been involved in this conversation for the past year plus what should they take from your results?

 

Morgan [00:41:25] So I would say first that those concerns are warranted. And certainly we should acknowledge those concerns in full, right, because we want to make sure that we don't do policing as a panacea. We think of it as one particular policy lever that is of interest. And there are certain things that we have to hold constant. Right. So there are certain things that the quality, maybe certain aspects of quality of policing and also managerial kind of directives and in structure, those are all things that matter deeply in terms of how police forces operate. And certainly one thing I think that the literature overall is starting to do is that they're starting to kind of open up that black box for policing and being able to figure out exactly what's going on. To the extent that policymakers are kind of collaborating with more researchers and academics, being able to kind of give us the ability to kind of work collaboratively on using more granular data to answer some very important questions about bias and other things, we should definitely do that. And it's certainly something that should be a priority for policymakers.

 

Morgan [00:42:24] What I will say is we don't necessarily kind of see that these benefits are always kind of shared across different types of neighborhoods. And in the director's cut, so to speak, of our paper, we did kind of look at this across the percent Black population distribution and kind of notice that the top quartile, we're not necessarily observing many of the benefits associated with increasing police force size, that is homicide victimization reduction. But we are still seeing many of the quote unquote "costs." That is more index arrests, still more index crimes. In addition to that, we also still see more quality of life arrests as well. And so to that extent, there still needs to be a very important, intense debate about how policing practices are carried out within Black communities throughout the country. I think about policing that kind of took place within Ferguson, Missouri, and the DOJ support around Ferguson. And so we don't want the police to have an increased presence and also to be using traffic violations as say, as revenue. That doesn't necessarily that's not within the realms of what we think of when it comes down to improving public safety.

 

Morgan [00:43:34] One thing that we also want to consider is that to the extent that there's an increase in civilian and police encounters, there's always the possibility that there could be use of force implemented as well. And so we definitely don't want that. And if we kind of use one of my coauthor's estimates that about 2.5 percent of arrests lead to police use of force, that our estimate suggests that police expansion necessary to abate one homicide, probably to about 7-10 use of force incidents, of which about 4-5 would involve a Black civilian. So we want to be very careful in kind of how we think about the police's role beyond the effects that they might have on homicide victimization and other considerations with respect to law enforcement activity.

 

Jennifer [00:44:16] Yeah, just to build a little bit on what you were talking about with the black box of policing, which I agree has been I think we're just getting to the point where over the last few years where researchers have had access to the kind of data we would need to be able to talk about what's inside that black box. And but I think the public conversations about policing often assume that we have to sort of take or leave this box. Right. We get this black box of policing and we either have more of it or we have less of it. And those are our options. And I think what I certainly would love, and it sounds like you would like also is for the conversation to shift more toward can we improve policing like there are these benefits that come from police. There are also these potential costs and are there ways to get the benefits with fewer of the costs? And that's something we just don't know yet. But it feels to me like we should be able to, like it feels like everything can be improved, right?

 

Morgan [00:45:05] Absolutely, and I think that a lot of times we hear these larger arguments about always incrementalism and incrementalism. Well, unfortunately, I think increments are just kind of within the reach of a public policy. If it were the case that we can improve neighborhoods through these kind of larger investments and other types of social services, and we knew this was evidence then I would be well behind it as well. And I think that we should continue to examine the role of other agencies and other interventions, because what again, the police are not a panacea. However, we don't know that experiment. We have not seen that experiment of what exactly happens if we reallocate a large and significant amount of funding away from the police to some other agencies. That requires a very deep monetary commitment. It also requires a deep transformation of the police and the agencies in question. And so I don't necessarily see it as a straight forward conversation, but I'm hoping that many of us will be kind of more enlightened about these conversations. That doctorate students might be listening to this and saying you know what, I have an idea that Morgan isn't thinking about. And more power to them, I can't wait to see the working paper.

 

Morgan [00:46:10] But I think that in many ways we have to be very careful of how we think about the role of police and whether we should scale it back or not, because at the end of the day, we do find important evidence that the police seem to have an important role through increases in labor and reducing homicide victimization disparity, something that is of the utmost importance to all of us. And so to the extent that pulling back the police in many parts of the country could lead to increases in homicide, did not want to necessarily be a good thing. And in many ways, as Aaron pointed out in his work, as I point out in my own work on gun control policy, places like St. Louis, places like Baltimore and elsewhere, I think that they would actually be arguing that they just there're just not enough policing resources. And I don't think this is just a super foolish union complaint. I actually think it's something that is actually real. You compare that to the levels of gun violence that might be existing in some of those same neighborhoods, they might have a point. So that's something that we want to consider.

 

Morgan [00:47:04] I think the other thing that I will say is that there's an important divergence sometimes in what the police do and what the community needs and wants. And so that divergence in the perceived public safety needs of Black communities in particular is something that has been a thorn within the larger discussion of law enforcement reform. In fact, in one of my- another book that I really enjoyed was Black Silent Majority by Michael Fortner at the CUNY Graduate Center. And so he kind of writes about the experience within Harlem, about dealing with the heroin epidemic in the late 20th century. And so there seemed to have been some very important differences between how the community thought that many of these kind of increases in crime, increases in drug utilization should be addressed and eventually how the state got involved and to the extent that this it started to spill over into suburbs, are people worried that will spill over into the suburbs, the state stepped in and we ended up kind of increasing the reach of the incarceral state going back to the early 1970s. And so these are very cautionary tales and it's very important to kind of acknowledge the role of the community and being able to kind of shape these larger law enforcement forces. And I think that sometimes at some point, that can kind of get lost within the larger literature. So I would be remiss if I didn't bring that up as well.

 

Jennifer [00:48:15] Yeah, so I know this paper is pretty new, but have any other papers related to this topic come out since you first started working on the study?

 

Morgan [00:48:22] Yeah. So I mean, there's been some work that hasn't necessarily been officially put out in the ether yet, so I won't necessarily comment on. And other work by maybe some rockstars by the name of Jennifer Doleac, Amanda Agan, and Anna Harvey has also commented on the role of some of these kind of low level offenses in the prosecution of those low level offenses. And I think that's kind of contributing to this larger conversation about how do we kind of go about disordered policing and prosecution now that we know a little bit more about their effects. Right. So we've had 20, 30 plus years of kind of evidence to kind of speak to how useful are these things. And I bought up the Stop, Question, and Frisk and the Floyd ruling case earlier. These are things that I feel as if are pointing us in the right direction about how to kind of shape policing as we move forward.

 

Jennifer [00:49:10] And what's the research frontier? What are the next big questions in this area that you and others will be thinking about going forward?

 

Morgan [00:49:16] So, I mean, I think a big kind of consideration, again, is more so kind of delving into the nature in which kind of racism is kind of driving some of the things that we see. Right. Again, I kind of call this a quasi experimental reduced form, non reduced form paper in the sense that, you know, there's certain things that we could certainly point to in terms of policy levers and being police employment. But there are obviously a lot of things that I mentioned before that that we need to continue to think about. Right. And so whether the police are kind of strategically kind of allocating their resources to Black communities, let's just say that when we speak about crime rates, we don't necessarily know exactly what a crime rate is many ways. We know what homicide rates for sure. But when it comes down to crime rates, there's a credible amount of imprecision which we're able to speak to these things because for some offenses it's going to be somewhat driven by the demand for policing and also the policing response. And so as a result, you know, maybe we use the homicide victimization rates across Blacks and whites as a signal. Maybe it just ends up being the case that police chiefs receive a lot of pressure to address crime in a significant way and particularly homicide. So they use homicide as a signal and they allocate resources that way. We don't necessarily know if that's true or not, but it is something we definitely need to explore, because, again, when it comes down to kind of understanding the nature of policing and exactly what type of policing different communities need, we have to have a finer understanding of perhaps the theoretical foundations and mechanisms that might be responsible for what- the equilibrium responses in policing that we're seeing at the end of the day.

 

Jennifer [00:50:54] My guest this week has been Morgan Williams, Jr. from Barnard College. Morgan, thanks so much for doing this.

 

Morgan [00:50:59] Thanks again, Jen.

 

Jennifer [00:51:00] You can find links to all the research we discussed today on our website, probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show. And thanks also to our Patreon subscribers. This show is listener supported, so if you enjoy the podcast, then please consider contributing via Patreon. You can find a link on our website. Our sound engineer is Jon Keur with production assistance from Haley Grieshaber. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening, and I'll talk to you in two weeks.

Jennifer DoleacInstrumental-Variables, Policing