Probable Causation

View Original

Episode 97: Allison Stashko

Jennifer Doleac

allison stashko

Allison Stashko is an Assistant Professor of Quantitative Theory and Methods at Emory University.

Date: July 4, 2023

A transcript of this episode is available here.

See this content in the original post

Episode Details:

In this episode, we discuss Prof. Stashko’s work on prosecutor elections and policy accountability:

“Prosecutor Elections and Police Killings” by Allison Stashko and Haritz Garro.

OTHER RESEARCH WE DISCUSS IN THIS EPISODE:

transcript of this episode:

Jennifer [00:00:08] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an economics professor and the director of the Justice Tech Lab. My guest this week is Allison Stashko. Allison is an assistant professor of business at the University of Utah, but this fall, she's joining Emory University in the Quantitative Theory and Methods Department. Allison, welcome to the show.

Allison [00:00:31] Thank you so much for having me.

Jennifer [00:00:33] Today, we're going to talk about your research on how prosecutors turn over affects lethal force by police, but before we get into that, could you tell us about your research expertise and how you became interested in this topic?

Allison [00:00:44] Sure. So I'm an economist and I study politics and crime in the US. So the way that this project got started is back in 2020, so if you remember, after George Floyd was killed and the sweeping protests and a renewed attention on prosecutors and police, the county attorney or the D.A. who was initially in charge of the case, brought the charge of second degree murder. And there was a lot of outrage over this decision and other decisions by the prosecutor's office before it was ultimately passed on to the state attorney general. So, you know, again, this brought up an old debate about who should handle cases when a police officer has killed a civilian and whether or not it's appropriate for the local prosecutor or the D.A. to handle those cases, which is a hugely important question in law. It's an important justice question, but as economists, we kind of sat back and thought, wow, this is also creating really bad incentives, and incentives can affect behavior.

Allison [00:01:45] So what we thought about was that fact that an officer maybe anticipates that there's a really low chance they're actually going to be indicted for a crime if, you know, in the worst case scenario, but that actually could also affect their behavior beforehand and so we wondered if that was the case and we started to think about how could we tease it out in the data by thinking about, well, what happens when a new prosecutor is in town and sort of wipe the slate clean where there's no longer this close working relationship between the police in a given agency and the prosecutor that they've gotten to know really well. And so we wondered if that would affect police behavior in such a way that they might also use less force.

Jennifer [00:02:24] Your paper is titled "Prosecutor Elections and Police Killings", and it's coauthored with Haritz Garro. So as you said, there have been lots of public conversations about the extent to which use of force by police, especially lethal use of force, is unnecessary. And whether increasing accountability for unnecessary use of force could change police officers behavior. So you've already kind of explained a little bit why prosecutors might be relevant here. How much have they actively been part of this conversation? And is this in the conversations you've you've heard this mostly has most of it focused on the police, or are people actually trying to change policies around what prosecutors are doing, too?

Allison [00:03:04] Yeah, great question. So ultimately, in this broader effort to improve law enforcement and to minimize unnecessary use of force, there are a lot of actors and there are a lot of people who are responsible for holding law enforcement officers accountable. It includes prosecutors, but also their immediate supervisors, police chiefs, sheriffs, mayors. And even if you zoom out, voters are indirectly responsible for holding all of these agencies accountable. And so there is a question of why prosecutors and I think the reason that they receive so much attention is that in these terrible, high profile cases, it's ultimately the prosecutor's choice as to whether or not to charge a police officer for having committed a crime and so that's a really important responsibility. And maybe unlike some of these other mechanisms for accountability, where it's written into a contract or it's sort of institutionalized, prosecutors have a lot of discretion and so they really play an important role.

Allison [00:04:02] And we have to think about their position in the broader context of police accountability. And, you know, as we mentioned already, the concern here is really that police officers and prosecutors by nature, because they work together and rely on each other day to day in their operations this might be a particularly weak accountability mechanism where we expect prosecutors to investigate and prosecute officers that they already know pretty well.

Jennifer [00:04:27] So what have we previously known about the effects of increasing accountability in, you know, whatever way we can on policing outcomes?

Allison [00:04:36] So there's so much research happening in recent years, and accountability is kind of a big term for how can we think about bringing law enforcement agencies in line with the public's interest. So I'll focus kind of specifically as I can about work that's related to our question. And so one way to narrow it down is to think about how police use of force is affected by policies, and then even more specifically, how do changes in the legal setting in which police operate affect use of force. So there there's an interesting study by Dharmapala, McAdams and Rappaport, that's in the Journal of Law, Economics and Organization from 2022, so fairly recent. And what they show is that after police officers gain collective bargaining rights, so this is actually sheriffs in Florida that then there was an increase in the number of reports of violent police misconduct. So kind of similar to the ideas that we'll be discussing here where there's a change in the actual or even anticipated consequences of use of force in this case going in the other direction where officers perceive maybe greater or have greater protection under the law, and that there you do see use of force increasing.

Allison [00:05:47] And then there's also an interesting study by Jennings and Rubado in the Public Administration Review from 2017 and there they find that things like reporting requirements, so whether or not you have to report pointing a gun, not just shooting or firing your gun, that that kind of affects are associated with use of force as well. So these are two studies where, again, it's something about the legal setting which affects police behavior and ultimately changes their use of force.

Jennifer [00:06:17] So that's I agree this is a really active research area, but given how important it is, people might be surprised that, you know, you can name two studies that have happened before, so why don't we know more than we do? What makes this so difficult to study? Is it mostly a data challenge or mostly an identification challenge or both of those things?

Allison [00:06:37] Yeah. So we obviously care a lot about identification, meaning we don't want to just report correlations, but we really want to find out if changes in a prosecutor office affect police behavior. So we care about that, but I have to say the answer is data. It's data, data, data. So in this case, what we want to do is think across the whole country and compare police use of force before and after a lot of elections. So we need data that covers the whole nation. And there's just not a lot about police use of force or police misconduct that's available at that scale. So a lot of the great work being done on policing is using data from large agencies and not looking across the whole country. So because we need that, we're really only able to see certain outcomes. And one of those outcomes, which is of course also a very important one, is the number of people killed by police, so fatal use of force only and so we think of this as sort of the tip of the iceberg. We would like to observe other measures of police misconduct, but data is a big challenge there.

Allison [00:07:38] And the other thing I'll mention about data is that it's only fairly recent that we have information about the number of people killed by police in the United States. So we now know that the official government statistics, which have been collected for a longer time period, they were missing more than half of the incidents. And so what happened was, after the death of Michael Brown in Ferguson, activists and journalists recognized this deficit of information about the number of people killed by police. And they kind of stepped in and they started crowdsourcing information and they started scanning local news reports for incidents of police killings. And that's the information that we have now. So this is organizations like Fatal Encounters, Killed By Police, Mapping Police Violence and The Washington Post and The Guardian also collected this information. So I think it's just a hugely valuable effort from these organizations and from these independent researchers that's enabled this research and also lots of other research on police use of force.

Jennifer [00:08:39] Yeah, I also just plug for, you know, grad students who might be listening. You know, another challenge here, if we want to know what happens when we increase accountability for police in a variety of ways, not just through prosecutor turnover, you need to be trying stuff, right? We need to actually be changing accountability measures somewhere so that we can see what happens. And we just don't have much of that change in the past to look at. So so this is an era where I think we're where lots of places are trying new things and passing new laws. And so hopefully that will provide a lot of interesting natural experiments that will provide that that identification we need to distinguish correlation from causation. But right now we're sort of we're in early days, I think, in figuring out what kinds of changes in accountability can actually affect outcomes.

Allison [00:09:27] I think that's absolutely right. Yeah, you can only study what's happening.

Jennifer [00:09:32] Yeah. Okay. So in this paper you focus on the election of district attorneys as a natural experiment. So let's talk about those DA elections. How common is it for places to choose DA's the election in the first place rather than through some other process?

Allison [00:09:48] Yeah. So most of the DA's in the United States or district attorneys are elected. They're only appointed in three states. That's New Jersey, Connecticut and Alaska, but otherwise you have these local prosecutors or DAs who are up for election every 2 to 8 years. And you can think of having approximately one DA per county. So they usually cover a county, but sometimes it's groups of counties. So we end up with about 2,300 elections every cycle.

Jennifer [00:10:19] Okay. And you talked about this a little bit, but let's dig in some more. Why might police change their behavior when a new district attorney's elected?

Allison [00:10:29] Yeah. So the idea that really drew us to this study was the conflict of interest between prosecutors and police officers. So it's well known that they work together closely. And so the concern is that they develop a relationship or a level of trust, which means that it's unlikely for a police officer to face an indictment. Now, of course, charging decisions are complicated and also the charging of a police officer has its own kind of particular considerations, but people wonder if this conflict of interest is also a factor.

Allison [00:11:00] And so if it is, then that could also affect police behavior. And one way to observe that is to look at these elections, because if a new prosecutor wins, then that kind of marks the end of what might have been a protective relationship with the incumbent or former D.A. and it marks the beginning of a new relationship. And so it might take time for that trust to rebuild. And we wonder if police officer behavior changes in that period. So first possibility is this conflict of interest story.

Allison [00:11:30] But then once you start thinking about it, there could be other things going on. So it could be that there is no conflict of interest. So just put that aside for a moment just the fact that there's a new D.A. in the office brings a lot of uncertainty, and the uncertainty can go in both directions. So police officers are trying to figure out this new D.A. and the D.A. also doesn't have a lot of information about police officers. And so when people face, you know, all of a sudden a larger amount of uncertainty in their lives, they tend to behave more cautiously. So that could also be explaining a change in behavior and in particular, it could lead to a more cautious behavior or less use of force. And then the last thing is that in recent years there's been this wave of progressive prosecutors. So the political landscape for the DA elections has changed a lot. And some prosecutors who are more reform minded are even campaigning on promises of holding police accountable. So you can imagine it that if the new D.A. is progressive and takes those positions, it would also have a different or separate effect on police behavior. And police might start to change their use of force given the agenda of the D.A. So those are the three main ideas that we have in mind.

Allison [00:12:40] It could also be, though, that another mechanism is where the new D.A. has incentive to develop a relationship with the police force. And so maybe actually the newest D.A. is the least likely person to bring charges and so it could go in the other direction. And so going into this, we had the the intuition that brought us to the question, but then we see actually there's a big question of A, whether or not there will be in effect and B, which direction it might it might go in. The only other thing I'll say about mechanisms is that maybe there are other things that could be happening, but one mechanism we kind of exclude or we think is not plausible is that a police officer is in an intense situation, has to make a split second decision, and pauses to think about recent local elections. So that seems really implausible, but sometimes when people hear about the concept of this paper, they they jump to that kind of more mechanical explanation. But no, we don't think that police officers are keeping election outcomes front of mind when they're in really serious and tense or dangerous situations. We wonder instead if there's sort of a big sea change when a new day takes over that affects policing even in sort of split second decisions.

Jennifer [00:13:55] Yeah, and so is the story that you have in mind then more one where, you know, there's just a general attempt to to sort of de-escalate anything. And so you don't you never get near that that split second fearful decision or something else. What's the kind of what's the--.

Allison [00:14:11] It could be it could be there's a big question over, you know, how many abuse incidents of fatal use of force could have been de-escalated instead. So it could be as police officers adopt more of a de-escalation tactic or they could be avoiding situations because they don't feel like the DA has their back anymore and so it could be a lot of different things going on. All of these things are sort of going to be in the background and we're just first going to see, does the number of people killed by police change in response to turnover in the prosecutor's office? And then we can start to tease out and try to better understand which of these mechanisms might be explaining that effect?

Jennifer [00:14:48] Yeah, kind of gets to a broader question. That is is also a hot research topic right now like to what extent is this use of force responsive to incentives in the first place versus just.

Allison [00:15:00] Yeah, yeah.
Jennifer [00:15:00] A function of fear say, yeah, so so you're you're testing basically are people responding to incentives in some way.

Allison [00:15:08] Yeah, that's right. I think there is another idea is that there are bad apples and good apples and that that explains excessive use of force and I think more and more research is pointing to the fact that actually these decisions do respond to incentives and to policy changes and to the surroundings as well.

Jennifer [00:15:27] Yeah. Okay. So in order to measure the causal effect of electing a new D.A., you need a treatment group and a comparison group and you need these groups to be very similar to each other, except for the outcome of the election. So how common is D.A. turnover overall, and what are the typical differences between places that elect new DAs and those that don't?

Allison [00:15:48] Exactly. So that's the idea, but the problem is that D.A. turnover is rare. It's it's not a common thing. So in the period that we have information on elections, which is 2012 to 2020, a new D.A. takes over only 27% of the time. So most of the time the D.A. stays in office for a long, long time and we should think of these turnovers as a pretty infrequent event. And that's a challenge for us because we can't just compare all the places that get a new D.A. to all the places that reelect the same incumbent D.A. and the reason is, it's not just that it's uncommon, but it's the places that tend to elect new D.As look different than the places that tend to reelect the incumbent over and over. So places that tend to elect a new D.A. are larger. So you can kind of think roughly where we're comparing a group that tends to be urban versus a group that tends to be rural when you look at new D.As and old D.As. And then also we can see that places that elect a new D.A. have more police killings per year leading up to the election and places that don't. So that's a real red flag where we don't want to just compare those places because we might get misleading results.

Allison [00:16:57] There's something about the places that tend to elect new D.As on average, which is different even with regards to the number of people killed by police than the places that tend to elect the incumbent D.A. and again, it's basically selecting more competitive districts and cities versus these rural smaller districts.

Jennifer [00:17:16] Yeah, there's probably a reason that places elected a new D.A. versus the places where they were fine with keeping the incumbent. And that reason, those underlying trends might it might be then difficult for you to tell if that was an outcome of the election or the cause of the election.

Allison [00:17:34] Exactly. These places that elect a new D.A., you wonder, well, was there a big controversy? Was there a high profile killing, which led to turnover not just in D.A's office, but in the mayor's office and the police chief and then you can't really attribute any of the changes after that to the election of the D.A. itself.

Jennifer [00:17:50] Right. Okay. So how do you start with that full group of elections where you worry a lot about, you know, what you being able to say? What's the effect of this election versus all the other stuff going on? How do you then narrow down that full sample to create the treatment and comparison groups that you need that are plausibly similar?

Allison [00:18:09] Yeah. So the trick is to focus only on places that had a very close election, right? And so within the set of places where the election was won by a very close vote margin, we can say that, well, there's really not going to be any big differences between places that elected a new D.A. by 52% of the vote versus places where the new D.A. or the challenger lost with a vote share of 48%. Why? Because those are such small changes that is really like at that point, it's a toss up and it's literally like a flip a coin when thinking about who won the election. And that's exactly the kind of outcome that we can work with to try to estimate causal effects, because it's not up to any of the candidates. They can't really control the votes with such precision to win in a very close election. It is a toss up. And so that way we can compare places that just narrowly had a new D.A. defeat the incumbent to the places where the incumbent just narrowly won.

Jennifer [00:19:06] Okay. So how exactly you've got this sort of coin flip idea where you comparing close elections and then at that point, is it as simple as just comparing places with where the the incumbent just barely lost with those that just barely won? Or tell us a little bit more about kind of exactly how you use these groups to measure the effects of electing a new D.A.?

Allison [00:19:29] Yeah. So we have the elections, which are, as best as we can tell, randomly assigning a district to having a new D.A. versus having the incumbent win reelection and we do see that in the data. So we end up with about 250 of these districts with close elections. And 50% of the time, the new D.A. does win a close election, which is really different from the full sample where only 27% of elections are won by new D.A. So we think we have our coin toss and now we want to compare what happens to the number of people killed in police in those districts where you have a new day in office versus in districts where the incumbent remains in office with one extra detail. So the districts where the incumbent wins is a good control group, but we want to make sure that the incumbent hasn't been in office for a short period of time. In other words, remember that these elections are ongoing in 2 to 4 years cycles. And so we specifically pay attention or we only use districts as a control. If they haven't re elected a new D.A. in the last eight years. So this incumbent has been around. There's no change or turnover in that office for some time and they're the control group versus the districts that happen to elect a new D.A. by a close margin.

Jennifer [00:20:50] And why do you restrict the sample in that way to have the longer time period for the incumbent being there?

Allison [00:20:56] Yeah. So because these elections are happening in different years, depending on what state you're in, you have to be careful, so you don't want to say that a district which elect a new D.A. in 2016 is a good control district the very next year when other districts are having their election. Why? Because we don't know what's happening in the post period after the election that's exactly what we want to find out. So as soon as a district has a new D.A. in our study period, or if they've had a new D.A. in the previous eight years, then we don't say that it's a good control district because they've had some turnover. So we're just using the places that have had a really stable D.A's office and comparing them to the places that do have changes. Just to kind of summarize what we end up doing, to estimate the effect is to compare how the number of police killings changes after an election versus before an election in a district that narrowly had a new D.A. win versus districts where the new D.A. narrowly lost. And I think listeners of your podcast will find this familiar. So this is the differences in differences strategy that we use a lot in economics.

Jennifer [00:22:08] Yes. And for people who follow the whole to a fixed effect, literature and economics and how complicated that difference in difference approach can be, you deal with all that stuff, too. So people who are interested in that can go check out the paper and we will gloss over it here. Okay, so let's move on to the data. So you pull a bunch of data sources together for this project. Tell us about all those data sources.

Allison [00:22:34] Yeah. So we have to stitch together four different data sources for this project. The first one, which I've already mentioned briefly, is the the number of people killed by police and there are a number of datasets available for this. These are all of the crowdsourced or open source datasets we use mapping police violence. We think this is a good one for our purposes because it includes both killings by officers who are on duty and off duty, which seems relevant for the mechanisms that we have in mind.

Allison [00:23:03] The other important piece of information is the outcomes of these D.A. elections. And so here another fantastic contribution comes from the Prosecutors and Politics project at UNC School of Law, and they collected all of the information about candidates and election outcomes for all D.A elections from the period of 2012 to 2017 and we've been working and also keeping this paper up to date, so we added election outcomes from 2018 to 2020. And this information is also coming from local news reporting or from official state records. And then the last two pieces of information come from the Census, so we want to be able to compare these districts that elect new D.As and that elect incumbents and make sure that our idea that this is a good natural experiment is actually supported by the data. So, for instance, from the Census, we can see that these districts, they are comparable in terms of population, demographics and socioeconomic factors. And then last, we bring in some other information from the FBI. So we look at the number of criminal offenses and arrests in a district and also the number of officers assaulted and killed. And we think these are important criminal justice outcomes to to tie into this analysis, so we'll also look at how prosecutor elections affect these outcomes.

Jennifer [00:24:21] Okay. So in this sample of places that had close elections, how often do police killings happen?

Allison [00:24:28] Yeah. So across the nation, there are about a thousand people killed by police every year or more. And so across the nation, also, 20% of districts have at least one person killed by police in a year. And in our sample of districts that have a close election is pretty similar. So 27% of districts have at least one person killed by police in the study period and the average number is 0.8 deaths per district per year. So it's not a very uncommon or unfrequent, infrequent event and so that's the main outcome of interest here.

Jennifer [00:25:03] Okay. And then how often are police officers indicted in those incidents?

Allison [00:25:08] The MPV or Mapping Police Violence does track whether or not an officer is indicted for any given incident, but it should be noted that it's it's not known how many of these incidents are tracked are following the number of indictments and so we don't know about the quality of this information as well as we do, but the actual number of police killings, but with that in mind, there are known charges for less than 1% of cases. And in our sample, it's about 2% of cases where there is at least one known charge. So, you know, in our sample, there's only 29 cases with an indictment over the years that we follow the district, which means that we're not actually going to be able to observe changes in the number of officers indicted, although that would be really interesting as part of the study. It's just such an uncommon event that we can't say much with any degree of precision with the data.

Jennifer [00:26:03] Yeah, and I guess we should also note, you know, it's not clear that the optimal number would be 100%. Right.

Allison [00:26:11] Correct.

Jennifer [00:26:11] So it's not clear that all of these police killings were unwarranted or unjustified, but to the extent that we think that police accountability is too low and police killings happen too often, we would expect more than 1 to 2% of incidents to lead to some sort of charge against the officer.

Allison [00:26:29] That's right. The correct number of cases that results in an indictment is is just unknown, but it is also true that, like most Americans in recent polls, agree that that there are too few charges or that officers are treated too leniently. So this is the sort of and the sentiment has been growing over time.

Jennifer [00:26:48] Yeah. Okay. So let's talk about the results. What was the effect of a new D.A's election on police killings?

Allison [00:26:55] Okay. So we do provide a range of estimates in the paper, but the the main result is that the number of people killed by police falls by 28% after a new D.A. ousts an incumbent. So that's a really large change in terms of a percent change. The baseline number of people killed in a district and in a given year is small though so this translates to about 0.2 deaths per district per year or you can think one death every five years.

Jennifer [00:27:24] Okay. And do those effects vary across the race of the civilian?

Allison [00:27:28] Yeah, that's a very important question. We know that the risk of being killed by police is much higher for black Americans and white Americans, and especially among young African-American men. However, in our study, while we see that those racial disparities in the outcomes, when we look at the changes after a new D.A. wins, it's pretty much the same across different groups. So if we separately estimate the effect for the number of people killed who are white, black and Hispanic, there's no statistical differences in these effects. And so it seems that the effect of a new D.A. on the number of people killed by police is not differentiated by the race of the civilian, as far as we can tell.

Jennifer [00:28:08] Okay. So just overall reducing the number of police killings and kind of benefiting everyone equally?

Allison [00:28:14] Yeah, there's just seems to be a leveling down of fatal use of force, but it's not specific to the race of the civilian.

Jennifer [00:28:20] Okay. And how long do these effects persist?

Allison [00:28:23] So the effects are really concentrated in the first year after the election. And in fact, you can think of this as like a one year shock. And then the districts that elect a new D.A. pretty much return to looking like the other districts that elected the incumbent. And so in the one year after the election, the effect is is much larger. It's like four deaths per district per year, which is almost a 50% reduction. And so it's useful to think of an average over all of the years, but when we break it down by the effect year by year after the election, you can see that it's actually driven by a short term reduction after the media is elected.

Jennifer [00:29:02] And we'll talk a little bit more about stuff you do around mechanisms in a bit, but that definitely seems consistent with your initial story that everyone's just like worried about what this new D.A. is going to do. Is that right?

Allison [00:29:14] Yeah, that's right. That's the that's consistent with the data and the way that I think of it what I think is most reasonable here is if you've ever had a new boss and you kind of straighten up on their first day of work and you go exactly by the rules, it might be a similar story here where there's a new person in a in a position of power or they have a huge responsibility and they could make an important decision in your life down the road if things go badly. So you play it safe while you figure them out and while they figure you out as well, but that's not really a permanent change. And so that story is consistent with the data so far.

Jennifer [00:29:49] Yeah. Okay. And you you also consider whether electing a new D.A. affects other policing outcomes, such as arrests and crime. So why are you interested in those outcomes? And what did you find?

Allison [00:30:01] Yeah, so this comes back to our earlier discussion also about mechanisms where it could be that police officers are using less fatal force because they're avoiding situations that would otherwise turn violent or they're able to de-escalate those situations. And so we don't have a perfect test. We don't have officer level data to really dig into this, but one thing we can do is ask if there are big changes in arrests in the districts. So if police are avoiding situations then or de-policing them, we would expect a decrease in arrests. And maybe following on top of that, you would see an increase in crime, but we don't see either of those changes after the election of a new D.A. So there's no evidence in a decrease of arrests or a substantial increase in criminal offenses, which leads us to say that it seems like police actually are de-escalating instead of de-policing.

Jennifer [00:30:54] Good news.

Allison [00:30:55] That is good news.

Jennifer [00:30:56] Yeah. And if police use less force against suspects in cases where that force would have been warranted, this could, of course, lead to more officer injuries or deaths. So de-escalating might actually put their own lives at risk. So you're able to test this directly. What do you find?

Allison [00:31:14] Yeah, so this is more good news. So we also think it's important to see if police are de-escalating are they taking on greater risk themselves? So from the FBI, we have the number of officers assaulted and killed while on duty and we see no increases in either of those measures. Now, these are also relatively rare events, and so we do not have as much statistical precision. We can't rule out a positive or negative effect or another way to say that we have a null effect, but we can rule out large increases. So, for example, we can say that if we put together the number of people killed by police and the number of officers killed, there is a reduction in the number of people killed. So we can rule out that there would be such an increase in the number of officers killed to cancel out the other effects and so this is why I say this is good news. We're not we are seeing a reduction in the number of people killed after a new D.A wins the election.

Jennifer [00:32:12] And with all the caveats that you put in before about how rare indictments are, you still check whether the election of a new D.A. increases the number of indictments of police officers. What do you find?

Allison [00:32:23] Yeah, so we actually can't see any change in the number of indictments because they are so few and far between, but what I will note is that even if we could measure precisely a change in indictments, we should note that it's not necessary to see an increase in indictments for the rest of the results to make sense. And I think this is an important point because we're really talking about how incentives change behaviors. And so it could be that police officers use lethal force less or they're like leveling down their use of force and de-escalating. And so then there would be fewer cases for prosecutors to investigate in the first place. So conceptually, it's not clear what should what we should expect with the number of indictments. And it's also true that prosecutors, yes, they make charging decisions, but they also make a lot of other decisions which are consequential for officers under investigations like the intensity of the investigation. And so for these reasons, we don't really have a clear idea of what to expect for the number of indictments. And then on top of that, there's just so few that we can't really say anything with quantitative analysis yet.

Jennifer [00:33:26] Yeah, this reminds me of the literature on police and crime, actually. So it's just another situation, right where you have most of the mechanism is through deterrence. And so you put more police on the streets and you see violent crime fall, but you often don't see arrests go up. And it's usually a very similar explanation, like a people are deterred because they now think the expected cost of this action has increased. Right. You've changed the incentives as you have here then you just don't do it anymore and so then there's nothing to arrest you for. And so so are police now are on better behavior because they're worried there is a higher risk of consequences with this new D.A. and there's nothing to indict them for. And so you wouldn't see a change in indictments. It's always tricky to kind of actually figure out then what the probability of consequences is when you don't see a change in that measure, but very similar story.

Allison [00:34:17] Yeah, that's that's exactly right.

Jennifer [00:34:19] Okay. And then you dig into the data a bit more to consider possible mechanisms that might be driving your main results. So tell us what the various mechanisms are that you're able to test and what you find.

Allison [00:34:31] Yeah. So the big thing we want to know is, is this a story of progressive prosecutors? So is it the case that during this time period, reform minded or progressive prosecutors came in with a hard line on police accountability and there were sweeping changes to police behavior that's one possibility.

Allison [00:34:49] And that it only took effect if that knew day won and was progressive. So we can test that because sometimes prosecutors are partisan, so we know if they're Democrat or Republican. And then on top of that, we also kind of go back to the news reporting on the candidates and on the campaigns to identify if they list police accountability or other progressive measures while they're campaigning so that they support those measures, but we don't find strong evidence that the agenda of the prosecutors so either their partisan affiliation or their their position on progressive reforms is driving the effects and here I also will just list one caveat or maybe two. One is that we only have about 300 close elections. And so as we start to slice and dice the sample into smaller groups, things get noisier. We can't really say that the result is driven by progressive prosecutors, but we can say that the point estimate is a little bit larger in magnitude or the effect is a little bit larger in those places that where the new D.A. wins and where they are a Democrat or progressive. So it's suggestive because it's not statistically significant, but it goes in the direction that you expect.

Allison [00:36:03] What I think is more interesting is that there is still a negative and statistically significant effect of D.A. turnover, even in districts where the new D.A. is Republican and in districts where there is no change in partisanship of the prosecutor. So from that, we conclude that agenda might be playing a role. It might be that the effects are bigger if you have a progressive prosecutor, but it's certainly not the only explanation. And then if we go back to mechanisms and you already mentioned one thing, is that because this effect is temporary, so it's really driven by changes in the first year and then disappears, but that seems to be in line with the relationship between prosecutors and police officers being broken when a new D.A. wins the election. But that relationship and trust developing again in a fairly short period of time.

Jennifer [00:36:56] Okay. So what are the policy implications of these results? What should policymakers and practitioners take away from your study?

Allison [00:37:03] Okay. So the one line summary of the study is that turnover in the prosecutor's office reduces the number of people killed by police, which, you know, I haven't mentioned it yet, but we were really surprised by this actually. Again, we thought that a fatal use of force is the tip of the iceberg and we weren't sure if we would see something there. So so that's big. And it's it's tempting then to just say, okay, well, then we need lots and lots of turnover and that's the policy implication. But there it's a little more nuanced because except for when prosecutors are investigating police officers, maybe it's a good thing that they have a very good, close working relationship. And so maybe you do want some stability in the prosecutor's office and you don't want lots and lots of turnover. There's also broader questions of prosecutor accountability. So maybe you don't want your prosecutor investing in lots and lots of elections rather than putting that effort into their job, maybe you don't want lots of lame duck prosecutor. So all of these kind of standard political economy concerns also could play out in the prosecutor's office. So for that reason, we don't say that clearly we want term limits or clearly we want shorter terms. That's all a bit complicated, but it's good to know that this could be a benefit of those changes.

Allison [00:38:18] On the other hand, I think a really clear policy implication is that it seems to be a very good thing for police officers to not know the person who's going to investigate and make charging decisions if they are under investigation. And that's something you can implement without making big changes to the legal system. So that's something you can implement by having a commitment to outside investigators and outside prosecutors. And that's a policy that some states are starting to pick up and so there you know, the take away of the study is is is relevant. We are extrapolating a little bit because that's not this be policy that we study in this work, but it seems plausible. It seems like that's something that definitely researchers should look out for as these policies do get implemented. So there we're back to the limitation where you can only study what's been implemented in some sense, but if they roll out to enough jurisdictions and enough states, I think that would be really interesting to see if the commitment to using outside prosecutors has a similar effect on police skills.

Jennifer [00:39:23] Yeah, I completely agree. I'll be keeping an eye out for for those policies and research on those policies. Have any other papers related to this topic come out since you all first started working on the study?

Allison [00:39:35] Yeah. So there's a lot of interest in now, like the agenda of prosecutors. It's no longer obvious that you just campaign by saying you're going to be tough on crime. There's the you know, the public sentiment has shifted. And so, of course, you and Amanda Agan and Anna Harvey have work looking at the agenda of progressive prosecutors and how it affects crime. There's also interesting work by Ashna Arora looking at the politics of prosecutors and incarceration. Dvir Yogev at UC Berkeley is looking at prosecutor agenda and jail population and Sidak Yntiso at Rochester is looking at partisanship of prosecutors again and racial discrepancies in charging decisions. So these are all zooming in to this important position of the prosecutor and thinking about other spillover effects in the criminal justice system, which I think is really interesting and important work as well.

Jennifer [00:40:27] And what's the research frontier, aside from studying these these independent prosecutor policies, what are the next big questions in this area that you and others are going to be thinking about going forward?

Allison [00:40:37] Yeah, it's such a good question and I wish we could phone in and hear what the audience thinks, but it's not that. But you know, another takeaway from this is that even as we are in a world where there are very few indictments and convictions of police officers, prosecutors have an important role in police accountability and they're affecting police behavior. And so I think if we zoom out, a next frontier is thinking about prosecutor accountability and who are the voters that hold prosecutors accountable? Who's participating in these elections? Whose interests do prosecutors ultimately represent? What's the role of media in prosecutor elections? Who are the candidates that are selecting into these campaign contests? So there's a huge number of, I think, sort of what would be standard political science and political economy questions that are open in this space of prosecutor elections and so that should be really interesting future work.

Jennifer [00:41:33] My guest today has been Allison Stashko from the University of Utah moving this summer to Emory University. Allison, thank you so much for talking with me.

Allison [00:41:41] Thank you.

Jennifer [00:41:47] You can find links to all the research we discussed today on our website probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show and thanks also to our Patreon subscribers and other contributors. Probable Causation is produced by Doleac Initiatives, a 501(c)3 nonprofit, so all contributions are tax deductible. If you enjoy the podcast, please consider supporting us via Patreon or with a one time donation on our website. Please also consider leaving us a rating and review on Apple Podcasts. This helps others find the show, which we very much appreciate. Our sound engineer is Jon Keur with production assistance from Nefertari Elshiekh. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.