Episode 46: Anna Harvey

 
Ep46_Harvey_photo.jpg

Anna Harvey

Anna Harvey is a Professor of Politics and Director of the Public Safety Lab at New York University.

Date: March 2, 2021

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

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Prof. Harvey's work on how court-ordered affirmative action in police departments affected racial disparities in crime victimization:

“Reducing Racial Disparities in Crime Victimization: Evidence from Employment Discrimination Litigation” by Anna Harvey and Taylor Mattia.

 

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 at 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 Anna Harvey. Anna is a Professor of Politics at New York University and Director of the Public Safety Lab there. Anna, welcome to the show.

 

Anna [00:00:27] Well, thank you so much for having me.

 

Jennifer [00:00:29] Today, we're going to talk about your research on how court mandated affirmative action in police departments affected racial disparities in crime victimization. But before we get into that, could you tell us about your research expertise and how you became interested in this topic?

 

Anna [00:00:44] Sure, I'd be happy to. So I'm an empirical social scientist by training, and I've worked on lots of different topics over the course of my career. About four years ago, I started to get interested in the question of racial disparities in policing. I was becoming increasingly troubled by stories that I was reading about the way that Black communities are policed. And so I started reading academic papers on racial disparities in policing. And I found some really careful and important papers on racial disparities in the way that police treat the suspected perpetrators of crime. So, papers looking at racial disparities in stops, citations, arrests, use of force, but I couldn't find any work about whether the police provide racially disparate protection to the victims of crime. And that seemed like a really important question because the justification for having police in the first place is that they protect us from becoming victims of crime. And if they're doing that in a racially disparate manner, if they're providing protection in a racially disparate manner, we should know that. But we don't have any work that looks at that question.

 

Anna [00:01:54] At the same time that I was reading those papers, I was also interested in and concerned about the way that police departments treat Black officers. So we know from lots of different contexts that organizations that discriminate or that tolerate discrimination in one area often also discriminate or tolerate discrimination in another area. So agencies that practice or that tolerate discrimination against Black officers may also practice or tolerate discrimination against Black victims of crime. And so any interventions that reduce discrimination against Black officers in hiring or promotion may also reduce discrimination against Black crime victims. And so at the same time, I started reading academic papers that looked at the effects of judicial interventions to reduce employment discrimination against Black officers. But I couldn't find any papers that looked at the effects of those interventions on racial disparities in crime victimization. So that's what gave rise to this paper. And I recruited one of our fantastic PhD students, Taylor Mattia, to work with me to see what we could contribute. And so in this paper, we're looking at whether court ordered interventions in police departments to reduce race based discrimination in employment also reduce disparities in the way that agencies treat Black victims of crime.

 

Jennifer [00:03:18] So your paper is titled, "Reducing Racial Disparities in Crime Victimization." And as you mentioned, it's coauthored with Taylor Mattia. So we're in the midst of a national conversation about how to improve trust between law enforcement and the communities they serve, and particularly between police officers and Black residents of those communities. But of course, this isn't the first time we're having this conversation as a country. This is a longstanding issue. So in your paper, you're considering the effects of court actions back in the 1970s and 1980s. So tell us about the context there and how the courts intervened.

 

Anna [00:03:51] Yeah. So during the 1970s and 1980s, many of the largest police departments in the country were sued for race based discrimination in employment and hiring. And some of these cases were brought by Black police officers as private plaintiffs and some were brought by the Department of Justice. And in many cases, private plaintiffs and the DOJ joined forces. So just as one example in Chicago, the litigation there started in 1970 after Black officers in the Chicago PD formed the Afro-American Patrolmen's League, which was a Black Officers Association. And the white leadership of the Chicago PD didn't want Black officers to organize. So in 1970, members of this Black Officers Association filed a complaint stating that they were being unconstitutionally subjected to adverse employment actions like charges or suspensions or dismissals because of their membership in the Afro-American Patrolmen's League. And by 1973, that complaint had expanded to cover hiring and promotion in the Chicago PD more generally, and it was joined by the Department of Justice. Now, in that case, there was a finding of racially discriminatory practices by the Chicago PD and there was an affirmative action plan in hiring and promotion that was imposed on the Chicago Police Department.

 

Anna [00:05:16] More generally, in 26 out of the 40 largest metropolitan statistical areas in the country in the 1970s and 1980s, law enforcement agencies were found to have engaged in race based employment discrimination and were subjected to a court imposed affirmative action plan in hiring or promotion. And so we're asking if those interventions to reduce racial discrimination in employment also reduce racial disparities in the way that safety was being provided to potential victims of crime.

 

Jennifer [00:05:51] So why might these court actions affect local crime rates or who's victimized by crime? What are the mechanisms you have in mind?

 

Anna [00:05:57] Well, police agencies tend to be whiter than the cities they police. So Black officer shares have persistently lagged Black population shares in our largest cities. And Black officers in many of these agencies have reported harassment and discrimination because of their race. So the hypothesis that we're working with here is that law enforcement agencies whose leaders practice or tolerate race based discrimination in employment may also practice or tolerate racially disparate responses to crime victimization. So we're thinking about in these agencies, officers may be allowed or even encouraged to, for example, take more time to respond to calls for service that are coming from Black neighborhoods relative to calls that are coming from whiter neighborhoods. To spend less effort trying to get Black victims to formally report their crimes to the police relative to white victims. To spend less effort trying to get Black victims and witnesses to testify relative to crimes committed against white victims. Even to spend less time patrolling Black neighborhoods as a way to deter crime relative to white neighborhoods. And if that kind of racially disparate response to crime victimization is happening, perpetrators of crimes involving Black victims may have lower probabilities of detection. They may be less likely to be caught. And so they may be less deterred from committing those crimes in the first place.

 

Anna [00:07:30] And so what we're wondering is if agencies are successfully sued over race based discrimination in employment and there's increased monitoring and scrutiny of their practices, they may also seek to reduce any observable and evident racial disparities in their responses to crime. So officers may be directed to decrease racial disparities in the response times to calls for service, in the rates at which victims kind of formally go on record to report crimes, the rates at which victims testify, crimes are cleared, the perpetrators are arrested and charged and convicted. And these efforts may increase the probability of detection of perpetrators who may commit crimes against Black victims. Over time, if agencies are successfully sued and these agencies are subjected to affirmative action plans that require them to hire more Black officers, that also may decrease Black crime victimization, if Black officers care more about detecting and deterring crime experienced by Black victims relative to white officers, or if, for example, Black officers have better information about the patterns of criminal behavior affecting Black victims relative to white officers.

 

Jennifer [00:08:42] So before this study, what did we know about what reduces crime victimization in general?

 

Anna [00:08:47] Well, you know, there are several really well known and really well done papers that show that hiring more police or deploying more police to a neighborhood does reduce crime on average. And those papers tell us that the marginal additional police officer hired under the city's police force or the marginal additional officers deployed to a neighborhood does in fact, on average provide protection to victims of crime. And these papers—you know the ones I'm talking about—they're really carefully done. I have a ton of respect for them. And I do think that just on a side note, these papers are really important to remember and grapple with in the current debate over police reform, that there is this evidence that police do provide safety on average.

 

Anna [00:09:35] But none of these papers looks at racial disparities in the provision of public safety or racial disparities in the extent to which we're protected from becoming victims of crime. That could be in part because of data issues. So most of the sources of data unreported crime, including the ones that have been used in those papers, don't tell you the race of the victim. And that includes the UCR crime data that are reported to the FBI by US law enforcement agencies. So to get victim rates, you have to use a source of data that reports victim race and one of those sources is the National Crime Victimization Survey that we're using in this paper. And if you look at the National Crime Victimization Survey, you do find persistent racial disparities in victimization over time, including within cities. So respondents were being policed by the same police force. Black civilians in that city are more likely to experience crime than white civilians. But we don't know anything about the nature of the police response to crime experienced by Black and white victims.

 

Anna [00:10:40] We also know a couple of things about them, the specific mechanisms by which police might reduce crime victimization. So one is about call response times. There's this great recent paper in the Review of Economic Studies on police response times in Manchester in the United Kingdom. And the paper uses 911 calls that are coming from locations that are close to each other, but that are located on opposite sides of a border that divides some neighborhood in the police district. And so as a consequence, these calls are coming from households that are geographically close to each other, but that have very different response times from the police stations that are tasked with responding to calls. And it turns out that calls that are coming from locations that are closer to response stations have shorter response times, but also have increased clearance rates and higher frequencies of identified suspects and higher frequency of immediate arrests. So we know that if you hire more police, you might be able to decrease response times. And that might be one of the mechanisms that's driving why police can provide public safety. We don't know anything about racial disparities in response times and clearance rates. We do know that the ACLU filed suit in Chicago in 2011 alleging that the Chicago PD had slower response times to calls coming from Black neighborhoods. And the ACLU for several years now has been trying to get the Chicago PD to provide data on its resource allocation across districts and response times, but they haven't been successful. So that's one mechanism that kind of might be driving what's going on.

 

Anna [00:12:19] Another mechanism that we know of that leads to hiring—you know, when you hire more police, you see reduced crime—is that we know that when you hire more police, you do get increased reporting of crimes. So Steve Levitt has this paper from a while ago where he found that increases in the size of a police force increase the ratio of reported non homicide crime to reported homicides—and homicides, we think are probably reported kind of almost all of them. And so the fact that that ratio reported non homicide crime as increasing as you hire more police suggests that part of what's happening is that more of the non homicide crime is being reported as you hire more police. So, again, we think that that may be one of the mechanisms, but we don't know if there are racial disparities in the responsiveness of crime reporting as you hire more police. And so it's kind of one of these open questions that were attempting to provide some evidence for.

 

Jennifer [00:13:19] And then what did we know about the types of court actions you're focusing on? So those aiming to increase the diversity of police forces, what do we know about how those affected police behavior and crime outcomes?

 

Anna [00:13:31] Well, there are two really great papers that have looked at the same kinds of judicial interventions that we're looking at. And both papers found that these interventions increased Black officer shares and reduced the Black representation gap or the difference between the percent Black of police employment and the percent Black of the population served. So there's one paper is by Justin McCrary and another is by Amalia Miller and Carmit Segal. And Justin also looked for effects of these judicial interventions on average reported crime rates. And he didn't find any effects. But reported crime, and he was using UCR data, reported crime combines both reporting—which might increase with increased police or increased effort by police—and victimization, which might decrease with increased police effort. And so- and those two things might offset each other so that you might find no observable effect of an intervention that affects police behavior. When what's really happening is that there are two effects. One is on victimization and one is on reporting. And those effects are moving in two different directions.

 

Anna [00:14:46] There is this great paper, and it's also by Carmit Segal and Amalia Miller, looking at litigation leading to gender based affirmative action plans in law enforcement. And they found that those litigation efforts increased female officer shares. They were also able to break apart reporting and victimization, and they found that increasing female officer shares as a result of these litigation efforts, increased rates of reporting of violence against women, rates of reporting to law enforcement, but also decreased the actual incidence of violence against women. So they were able to distinguish those two offsetting trends. And that's essentially what we're trying to do here. But for crimes committed against Black victims.

 

Jennifer [00:15:35] So this is a tough question to study for a few reasons. So what are the hurdles that researchers like yourselves have to overcome in order to measure the causal effects on crime victimization? Are these mostly data challenges or identification challenges or is it both?

 

Anna [00:15:51] Well, you know, it's always both. It's always everything. So the first hurdle is that you have to have a source of data that identifies the race of crime victims. And so that basically rules out, not exclusively, but it basically rules out the UCR data. The National Crime Victimization Survey does identify victim race, but the victimization survey doesn't provide geographically disaggregated data because of concerns about confidentiality. Except that there was one release of victimization survey data between 1979 and 2004 for the 40 largest MSAs or metropolitan statistical areas. So this data release identifies the MSA of the respondent in the survey, and those are the data we're using in this paper. And that's part of why we're focusing- that is the reason why we're focusing on the time period that we're focusing on. It was the only data that we have.

 

Anna [00:16:47] The second hurdle, though, is that, as you said, you have to think really hard about what it takes to make a causal claim about the effects of reducing racial discrimination in employment, on racial disparities in the police response to victimization. We'd like to know if agencies that reduce racial discrimination in employment also reduce racial discrimination in policing. But imagine that we just collected data on, say, Black officers shares, maybe as an indicator of how welcoming a department was to Black officers. And we collected data on crime victimization by race and we asked whether agencies that hire more Black officers also reduce racial disparities in crime victimization. It would be really hard to make a causal inference there because agencies that hire more Black officers might be different on a number of dimensions relative to agencies that hire fewer Black officers. So they might be located in cities that have more liberal voters or city councils that are more actively involved in police oversight or have more active or involved community organizations that are advocating for racial equity in policing.

 

Anna [00:18:02] So any of those factors might lead a police department to seek to both increase Black officers shares and reduce racial disparities in crime victimization without the former causing the latter. So you need something external to police departments ideally, right, that causes them to act specifically to reduce race based discrimination employment and isolation from these other potential causes as much as you can. And then you could look at the effect of those actions on racial disparities in victimization.

 

Jennifer [00:18:35] So you're going to use the timing of litigation against police departments for exactly this purpose to get traction on this question. So let's think through, I guess, what the ideal experiment is that a researcher might like to run in this context and how your approach approximates that experiment.

 

Anna [00:18:52] Right. So what we want to do is we want to isolate some cause, external to police departments, that causes them to reduce racially discriminatory employment practices. And we're using litigation leading to judicial interventions that direct agencies to end discriminatory practices and hire more Black officers. And ideally, what we'd have is some kind of field experiment where we could randomly assign agencies to either receive this intervention or not receive it. Right. So we'd pick agencies out of a hat and some of them would get judicially supported and monitored orders to end discriminatory practices and hire more Black officers and others wouldn't. But we can't run that experiment for obvious ethical and feasibility concerns. And the agencies that were litigated for race based employment discrimination and were subject to these judicial interventions might have been different on a number of dimensions relative to agencies that weren't litigated. And those differences might have affected racial disparities in crime victimization. Our strategy doesn't really get us entirely out of the woods.

 

Anna [00:20:04] One thing we can is to use just the set of MSAs that contain law enforcement agencies that were treated or that were actually litigated for race based employment discrimination and subjected to these interventions. All of these agencies share the important similarity that their big city departments that were found guilty of race based discrimination employment in law enforcement agencies and that were ordered to hire more Black officers and end discriminatory practices. So they all share that important commonality. But the timing of these litigation efforts varied across departments. So we can maybe use variation in the timing of litigation onset between 1970 and the end of the 1980s as a source of variation in treatment. And so we explore this empirically in the paper, and it turns out that you can actually predict pretty well which agencies will be treated, meaning they'll be litigated and subjected to these interventions if you use census data from 1970 before any of these litigation efforts happened. But you can't predict the timing. If you just isolate the set of agencies that did get these interventions, we aren't able to predict when those interventions were happening. So that's what we're leaning on in the paper, is that the timing was at least- couldn't be anticipated given what we know about pretreatment data.

 

Jennifer [00:21:32] Yeah. So the intuition here then is it's essentially random, conditional on having litigation against you at some point. It's essentially random when that happens. Then maybe there are just sort of underlying differences in getting the case together or whatever, but at least it's not going to be related to those trends and victimization rates. Am I getting that right?

 

Anna [00:21:53] That's the idea. And it's not perfect. It is a strategy that we've seen used in other contexts. In, for example, people looking at the effects of school desegregation litigation or people looking at the effects of, say, disability insurance offices that close- you know, in all of those cases that the quote unquote "treated" agencies or offices are often distinguishable from the offices or agencies that aren't treated. But within the set of treated agencies or offices, the timing looks like it can't be predicted. And so, right, that's what we're leaning on.

 

Jennifer [00:22:28] Yeah, and then natural experiments are never perfect, unfortunately. They might be close, close to the ideal experiment. OK, so what data are you using for all of this?

 

Anna [00:22:38] So we're using a litigation database that was put together and really graciously made available by Amalia Miller and Carmit Segal. So they searched federal litigation databases for employment discrimination cases involving law enforcement agencies that had at least two hundred protection employees. And then they further identified cases in the set that resulted in court orders or settlement agreements that imposed affirmative action plans in hiring or promotion on those agencies. So we're pulling from that database litigation data for any county or municipal law enforcement agencies that are located in one of the 40 largest MSAs that are in the National Crime Victimization Survey sample. And there are 167 of those agencies. And so we're putting together the litigation data with the National Crime Victimization Survey.

 

Jennifer [00:23:32] And so tell us more about that survey. Who's in that sample and what questions are they asked?

 

Anna [00:23:38] Well, so the NCVS, it's been conducted every year since 1973 by the US Census Bureau on behalf of the Bureau of Justice Statistics. It's a pretty high quality survey. It uses a nationally representative sample of about 50,000 housing units. And once a household is sampled into the survey, they stay in for three years. And members of the households are interviewed every six months about whether they've experienced any incident of crime over the previous six month period. And they were specifically prompted about whether they've experienced anything like robbery, burglary, theft, assault or rape. And if they answer yes to one of these initial prompts, then there are additional follow up questions that get more details about the nature of the incidents, including whether it was reported to the police. And there are some questions about the nature of the police response. And the NCVS also, importantly, collects demographic information about respondents.

 

Jennifer [00:24:43] So you mentioned earlier that there are long standing gaps, racial disparities in victimization rates. So what are the baseline victimization rates for white and Black respondents of the survey before litigation occurs?

 

Anna [00:24:57] Well so, if we just look at what we're reporting in our main results, which is just the sample of treated MSAs—so these are MSAs that all have law enforcement agencies that would eventually be subjected to litigation leading to race based affirmative action—prior to litigation, on average victimization rates are 20 percent for Black respondents and 13 percent for non Hispanic white respondents.

 

Jennifer [00:25:24] OK, so let's get into the results. So you first use an event study model to measure the effects of litigation on crime victimization by race. So visually here you're going to graph the effects over time. And the idea is if the litigation mattered, we should see a change in the trend at the time the litigation occurs. This is a very visual design. We- it's always hard to talk about these over a podcast without the graph in front of you, but let's do the best we can. So what do you find when you do this?

 

Anna [00:25:53] Well, we find that prior to litigation, in the years leading up to litigation, we're not seeing any trends in either white or Black crime victimization. But then after litigation, we see decreases in both white and Black victimization relative to the baseline pretreatment rate. But these decreases in victimization are much larger for Black respondents. So white respondents see, on average, about a four percentage point reduction in crime victimization after litigation onset. But Black respondents see about on average a 10 percentage point reduction in crime victimization following litigation onset. And the other thing we find is that these, what we call treatment effects or the effects of litigation, appear immediately for Black respondents and they gradually grow in magnitude. And so this is- and whereas for white respondents, it takes a little more time for these effects to appear. So what's happening here is we're seeing an immediate change in litigated agencies responses to reports of Black victimization, followed by gradually increasing reductions in Black victimization that are possibly caused by gradually increasing Black officer shares.

 

Jennifer [00:27:13] And then what's the effect on the racial gap in crime victimization?

 

Anna [00:27:17] Right. So we directly model the racial disparities in victimization rates and we see substantial reductions in those disparities. And so our estimates of these decreases in racial disparities range between six and nine percentage points over the 25 years after litigation onset. So that's in every year relative to the baseline year. That's kind of the average decrease that we see relative to the pretreatment baseline year. And these estimates are all significant at the 95 percent threshold. On average, there's about like a seven and a half percentage point decrease in the relative Black victimization rate. And remember, I just told you a few minutes ago that prior to litigation, there was about a seven percentage point gap in these MSAs in Black and white victimization. So Black respondents had a victimization rate of about 20 percent. And white respondents had a victimization rate of about 13 percent. What we're seeing is that in these cities that were litigated, that gap in victimization is almost entirely eliminated after litigation.

 

Jennifer [00:28:29] So these results are amazing and huge. So then you start thinking about what's driving them, what are the potential mechanisms? And the first thing you look at is whether there's any change in whether crime is reported to police. So what do you find is the effect of litigation on reporting rates?

 

Anna [00:28:46] So. Right. So remember, one of the things we talked about, why why do the police—or I guess maybe it's how—how do the police protect us? One reason is that we, you know, when you hire more police, you see more reporting of crime to police. So most of the crime we experience, we don't actually report to the police. But you see a larger fraction of that crime reported after more police are hired. What we're seeing in our data is that after litigation that forces agencies to reduce racial discrimination employment and to hire more Black officers, we also see increases in reporting rates for both white and Black respondents.

 

Jennifer [00:29:28] And does that effect differ by race at all?

 

Anna [00:29:31] It does. So we see increases for both white and Black respondents, but the effects are substantially larger for Black respondents. So white respondents see an average of about a two percentage point increase in reporting over the 25 year after period after litigation. But there's about a six percentage point increase in reporting for Black respondents. We have pretty large confidence intervals point estimates for Black respondents. And we can't reject the hypothesis that the increases for Black and white respondents aren't the same, but it is suggestive evidence, I guess I would say, that these increases are larger for Black respondents in terms of reporting crime to the police relative to white respondents.

 

Jennifer [00:30:20] And so you mentioned earlier that if victimization rates are going down, but reporting rates are going up, which is what you find, these effects might cancel each other out if you were just using reported crime data as the outcome. So you look at this directly, and what do you find when you just use reported crime as the outcome?

 

Anna [00:30:38] Yeah, and so remember, that Justin McCrary has a really great paper on the effects of these kinds of judicial interventions on reported crime rates. And he didn't find any effects. But he was using administrative data on crime reported to the police. And as you said, if this litigation was decreasing victimization but increasing reporting, then those trends might offset each other. And so we look at that and that is what we find in the victimization data, that the magnitudes almost exactly offset each other. Victimization that is reported to law enforcement in the NCVS is essentially unchanged after litigation for about 15 years after the onset of litigation leading to affirmative action. And then we see some very small decreases. And so we're using different data and a different sample of cities than Justin did. But we're essentially able to replicate the null result that he found in reported crime. What we're doing is we're showing that that null result is the product of real decreases in victimization and real increases in reporting,

 

Jennifer [00:31:46] Which then is in line with that Miller and Segal paper you mentioned on police officer gender, right. Where they were finding reductions in victimization of women and increase in reporting. So it all just sort of ties everything up with a nice bow. It's always nice when this happens. Right.

 

Anna [00:32:02] And Jen it is, and actually to put- to kind of underscore that point, in a different Justin McCrary paper one of the things he shows is that increases in police have a really large effect on homicides, but they don't have any effect on crimes like rape. And you have to be really careful when you're looking at results like that, because you might say, well, we should really direct law enforcement agencies to devote all of their attention to homicides because that's where the effect is, but not devote any attention to, say, sexual assault crime because they don't have any effect on sexual assault crime. But precisely what you're saying is in the Miller and Segal paper, they're showing that that null result for crimes against women is really the product of two offsetting trends.

 

Jennifer [00:32:47] Yeah. So finally, then you look at the reasons respondents give for not reporting crime in the cases where they don't report to police. So does litigation affect their stated reasons for not reporting?

 

Anna [00:33:01] Yeah. So the NCVS asks respondents who experienced a crime but who did not report the crime to the police, why they didn't report to the police. And they give us like the most frequent answers are called out in the victimization survey. So some of these reasons indicate a lack of trust in the police response. So the respondent thought that if they reported, the police wouldn't help or they wouldn't do anything. Other reasons for not reporting are unrelated to the nature of the police response. So, for example, maybe the respondent thought the crime just wasn't important enough to report or they dealt with it in another way. So we separate out those responses. We we look at those separately—responses that indicate that respondents didn't trust the police to respond if they reported and responses that are for some other reason.

 

Anna [00:33:51] And we find that after litigation, not reporting crime to the police because of a lack of trust in the police response, it decreases for both white and Black victims of crime. But these decreases are larger for Black victims. So the average post litigation decrease in not reporting victimization because of a lack of trust in the police response is about seven and a half percentage points for white respondents, but it's 13 percentage points for Black respondents. So these estimates suggest that Black respondents became more confident in the nature of the police response that they could expect after reporting after the onset of litigation leading to affirmative action. When we look at the reasons for not reporting crime the victimization are unrelated to trust in the police. We don't find any changes after litigation, either for white or for Black victims of crime.

 

Jennifer [00:34:56] So the goal of the litigation was to increase the number of Black officers on the local police force. And this could itself be a reason for the reduction in crime victimization in Black communities for a few reasons that you alluded to earlier. Did you find that litigation had its intended effect on the racial composition of those local police departments?

 

Anna [00:35:14] Yes, and our findings are totally consistent with the results that have been reported by Justin McCrary and by Miller and Segal. Over the course of the 25 years after the imposition of litigation leading to affirmative action, the average litigated agency saw about a five percentage point increase in Black officer shares. And if you look back to the kind of average baseline rate prior to litigation, these agencies had about 12 percent of their officer force comprised of Black officer shares. So we're seeing about a forty two percent increase in Black officers shares relative to that baseline rate.

 

Jennifer [00:35:54] Alright. And then you also consider a few other potential causal mechanisms for the finding that crime victimization is falling and you basically rule them all out. But what were some of the other mechanisms that you were thinking about and able to test?

 

Anna [00:36:09] Well, you know, because there are so many papers about the beneficial effects of hiring more police, we thought, well, maybe one of the mechanisms is just that this litigation causes you to hire more police and that's driving the results. So we looked at that, the effect of litigation on numbers of sworn officers per capita and we didn't find any effect. We also thought, well, maybe these these MSAs that are getting litigated and are subjected to these judicial interventions, maybe they're also experiencing demographic changes and maybe the litigation itself is causing demographic changes like just in the composition of the population. And we can't find anything there. And we also thought, well, maybe there are certain kinds of crime that are driving the effects that we're seeing, so we look for heterogeneous effects of litigation across different types of crimes. And we didn't find any patterns there.

 

Jennifer [00:37:00] So what's your overall takeaway about what drove the effects that you're seeing? Which mechanisms mattered most?

 

Anna [00:37:06] Well, if you just kind of step back and think about what we already knew about how police provide public safety, we know that marginal increases in police hiring increase safety and reduce crime overall on average. And we think that at least part of how this happens is both because we report more crime to police and because with more resources, the police improve the nature of their response to our calls. And the end result of that is that perpetrators of crime are more likely to be caught if they commit a crime. And in response, this is classic Gary Becker, crime goes down. And what our findings reveal is that where law enforcement agencies were engaging in discrimination against Black officers, Black civilians who weren't reporting victimization to the police were more likely to say that they weren't reporting because they didn't believe that the police would respond relative to white civilians. After these agencies were sued for discrimination against Black officers and forced to accept interventions that were designed to reduce racial discrimination in employment, Black civilians became more likely to trust in the police response relative to white civilians and Black civilians also suggestively became more likely to report victimization to the police relative to white civilians.

 

Anna [00:38:24] Now, really importantly, we don't have data on the actual police response to Black crime victimization. So if you called, did the police come? If so, how long did it take them to get there? What did they do when they got there? What was the race of the officers that responded? Was the offense cleared? Did they identify the perpetrator? Did they arrest somebody? Did they convict somebody? We don't have any of those data. All we have are these data from respondents. But the survey data are at least consistent with the notion that agencies improved their police response to Black victimization after they faced serious litigation for discrimination against Black officers. Some of this improvement happened very soon after litigation onset, suggesting that existing officers, for whatever reason, took Black victimization more seriously. And other improvement happened later during the time that Black officers shares were increasing. So I guess the overall takeaway is that external interventions to reduce race based employment discrimination in law enforcement agencies can also reduce both absolute and relative Black crime victimization without increasing white crime victimization.

 

Jennifer [00:39:38] Yeah, I do want to highlight for folks that—I mean, I found it just fascinating—this isn't all just coming from hiring more Black officers, right? I mean, I think they're- sort of the way most people, including me, kind of approach this issue is there's something about the composition, the racial diversity or gender diversity of the police force that gives the local community trust that their concerns will be heard and cared about, maybe if the officers look like me. But you're seeing results, you're seeing effects before that increase in hiring actually happened. So it really is something about being forced to address the racial discrimination in employment rather than the actual composition of the police force, which is something I really hadn't heard hypothesized before or seen studied before your paper. I think that's super interesting.

 

Anna [00:40:27] Yeah, I think that's right.

 

Jennifer [00:40:29] So, OK, so that's your paper. Have any other papers related to this topic come out since you first started working on the study?

 

Anna [00:40:36] Well, I'm so glad you asked that because just about a month ago we got a new paper by Aaron Chalfin and Ben Hansen, Emily Weisburst, and Morgan Williams looking at whether there are racial disparities in the effects of increases in police hiring on safety. There's one place in the UCR data where you can get race of victim, and that's in the supplemental homicide data. So for homicides, the FBI collects additional information about crimes, including the race of the victim. In all these prior papers about the beneficial effects of police hiring and safety, people had used the homicide data, but they hadn't used the supplemental homicide data that break down homicides by race of victim. And so Aaron and his coauthors are looking at whether there are racial disparities in increases in police hiring on homicide rates.

 

Anna [00:41:31] And they find that for cities that don't have large numbers of Black residents, increases in police hiring do reduce homicides of both Black and white victims. In the cities that have large numbers of Black residents in the top quartile of cities—and these are cities with Black populations of about 25 percent and larger, which includes many of our largest cities—increases in police hiring between, I think they're looking at like 1980 to the present, reduce homicides of white victims but have no effect on homicides of Black victims. And that just seems like a really important finding to me. We have all these papers on the beneficial effects of hiring more police, but nobody's ever looked at this question about really, I guess, the distributional effects on safety of hiring more police. So this team, they don't have an explanation for this finding. But just documenting the existence of these racial disparities in the provision of safety is I think really important.

 

Jennifer [00:42:32] Yeah, that's super interesting. So what are the policy implications of your results and the other work in this area? Which should policymakers be taking away from all of this?

 

Anna [00:42:40] Well, I think the big takeaway is that interventions that address employment discrimination against Black police officers, including but not limited to, forcing agencies to hire and promote Black officers, can also reduce crime victimization among Black victims. And so I think this is really important that improving the working conditions for Black officers is actually linked to improving how well police agencies do their job in providing protection to Black civilians. There are a number of large police departments right now that have pending litigation against them, filed by Black officers alleging discrimination, including my own police agency, the New York Police Department. And these litigation efforts, I think, are an indicator that these agencies are probably also providing safety in a racially discriminatory manner. So I think it's really important to think about how addressing workplace conditions for Black officers can also improve safety for Black civilians.

 

Jennifer [00:43:47] And then what's the research frontier? What are the next big questions in this area that you and others will be thinking about going forward?

 

Anna [00:43:54] The one thing we really lack in our paper is that we don't have any granular, real time data on police response to reports of crime made by Black and white civilians. So we just talked about this, when if you called, did the police come? And how long did it take them to get there? And what did they do when they get there? And who's responding? And was the offense cleared? And where a patrol car is spending their time. All of this really granular data. Police departments have all this data, and we need access to those data to be able to push further on these questions of how well and how equitably police are providing safety to their communities. You know, it's the job of police agencies to provide safety. And you think that agencies would care about how well they're doing their jobs, about how equitably they're doing their jobs. But very few police departments are willing to allow researchers to have access to their internal data to assess their performance. And I think that needs to change.

 

Anna [00:44:50] I will say a couple of things. I do think that there are some major city chiefs out there who care about performance and about equity in performance. A few years ago, the police chief of one of the largest cities in the country, one of the top five largest cities in the country, asked me to come in and look at how his department was doing in responding to calls for service and whether they can improve their performance, including the equity in their responses. But when the department's lawyers got wind of what we wanted to do, they raised all kinds of objections. And they stopped the study from going ahead because they were worried about what the study might reveal about the way the department did its job. And I just think that kind of attitude has to stop. One positive thing that we've seen over the past year in the debate over police reform is that some cities are voting to force their police departments to provide data to researchers. So you and I know that Charlotte, North Carolina, just did this. They voted to require an independent evaluation of a number of features of the way their police department is conducting business, including providing access to the data. And I think it would be a really positive thing if more city governments, when they're thinking about police reforms, would think about provisions requiring departments to be the subject of evaluations by university based researchers, including sharing all their data.

 

Jennifer [00:46:16] Here's hoping. That would be an amazing step forward. My guest today has been Anna Harvey from New York University. Anna, thanks so much for talking with me.

 

Anna [00:46:26] It was such a pleasure. Thanks for having me.

 

Jennifer [00:46:34] 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 John Keur with production assistance from Haley Grieshaber. Our music is by Werner, and our logo was designed by Carrie Throckmorton. Thanks for listening, and I'll talk to you in two weeks.

Jennifer DoleacPolicing, Difference-in-Differences