Episode 59: Kerri Raissian
Kerri Raissian
Kerri Raissian is an Associate Professor of Public Policy at the University of Connecticut.
Date: September 28, 2021
Bonus segment on Professor Raissian’s career path and life as a researcher.
A transcript of this episode is available here.
Episode Details:
In this episode, we discuss Prof. Raissian's work on domestic violence courts:
“Specialization in Criminal Courts: Decision Making, Recidivism, and Re-victimization in Domestic Violence Courts in Tennessee” by Aria Golestani, Emily Owens, and Kerri Raissian.
OTHER RESEARCH WE DISCUSS IN THIS EPISODE:
“Love, Hate and Murder: Commitment Devices in Violent Relationships” by Anna Aizer and Pedro Dal Bó.
“Do Domestic Violence Courts Work? A Meta-Analytic Review Examining Treatment and Study Quality” by Leticia Gutierrez, Julie Blais, and Guy Bourgon.
“The Impact of Specialized Prosecution on the Safety of Domestic Violence Victims” by Ashna Arora, Xander Beberman, Zubin Jelveh, and Ashley Motta. (Available upon request from the authors.)
“Misdemeanor Prosecution” by Amanda Agan, Jennifer Doleac, and Anna Harvey.
“Can Restorative Justice Conferencing Reduce Recidivism? Evidence From the Make-it-Right Program” by Yotam Shem-Tov, Steven Raphael, and Alissa Skog.
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. My guest this week is Kerri Raissian. Kerri is an associate professor of public policy at the University of Connecticut. Carrie, welcome to the show.
Kerri [00:00:27] Hi, Jen. Thanks so much for having me.
Jennifer [00:00:29] Today, we're going to talk about your research on domestic violence courts. But before we get into that, could you tell us about your research, expertise and how you became interested in this topic?
Kerri [00:00:39] Yeah, sure, I'll be happy to. So I actually started my career, my first job as a victim advocate know happily enough, like actually in the district attorney's office, which is going to be featured in this paper in Davidson County in Nashville, Tennessee. And as a as a victim advocate, I just wanted to help people leave dangerous relationships. I wanted to be a part of that process and I loved that job. It was very rewarding. But after about three or so years, I left to manage a domestic violence shelter, also in Nashville and I also left that job, which was also super rewarding. But over time, I realized that for every victim that I think we helped as advocates, I think there was another victim who the intervention might not have been ideal for or might not have helped at all. And so I really wanted to learn how to do policy analysis to determine if interventions and policies and all the things advocates that we hope work sort of really did work. And in particular, I also wanted to think about if there were unintended consequences to the things that we were that we were doing.
Kerri [00:01:46] And so I really consider my research expertise to be twofold. I have a pretty good sense of how the quote on quote system works, which has helped me identify quirks in the system that maybe we can use to exploit plausibly exogenous variation, right? And that makes for good research design, especially in this case, but I'm also trained in policy analysis and economic demography. So this paper is going to look at the effect of domestic violence court specialization on DV case outcomes and that was my very first job working in these courts. And I just have very much long wanted to know what the effect of these courts is for for victims and defendants. And so this paper is particularly satisfying in that respect.
Kerri [00:02:30] But also, I think it's a neat paper and I'm excited to talk about it. And it's just been really great thinking about this with Emily Owens, my coauthor at UCI and Aria Golestani who I should mention is an amazing Ph.D. candidate in economics at UCI. And he will be available for interviews during the 2022 ASSA meetings. He's on job the market, on the market, and so I'm really excited for him to see what's next.
Jennifer [00:02:57] Awesome. Yes, everyone interview Aria.
Jennifer [00:03:01] OK, so your paper is titled "Specialization in Criminal Courts: Decision Making Recidivism and Re-Victimization in Domestic Violence Courts in Tennessee," as you mentioned, it's coauthored with Aria Golestani and Emily Owens. So let's start with some basics. What our domestic violence courts and what makes them different from regular courts?
Kerri [00:03:20] Yeah. So I think we maybe also have to think about what makes domestic violence different from other kinds of crimes to think about what domestic violence courts are and what problem-solving courts are in particular.
Kerri [00:03:34] So these courts exist because domestic violence cases have particular elements that makes prosecuting them just really hard and makes deterring future crime or future domestic violence incidents very hard, right. Domestic violence cases are different from many crimes because the victim and the defendant know each other. They have an intimate relationship or familial relationship. They may live together, they may share children, they may share property and so forth, right? So it's complicated. And all of those things can mean that the victim faces additional barriers in separating from their abuser, they may not want to separate from their abuser, or they may not feel like they can separate from their abuser. Maybe they have separated from their abuser, but there is repeat abuse and harassment, so this cycle is really challenging for the victim and for prosecution.
Kerri [00:04:23] And in fact, these circumstances and dynamics that lead to the victim's reluctance or even refusal to prosecute, I think are summarized really nicely in Aizer and Dal Bo's 2009 "Love, Hate and Murder" paper. So they have a really nice model for thinking about this. But I think these characteristics have real implications for victims safety, which also means that the criminal justice system, I think, has a particular responsibility to victims to ensure their safety is maintained while charges are pending and of course, after charges get adjudicated right. So in summary, right like domestic violence cases are notoriously difficult to prosecute and so one particular way to deal with that is to to have these problem-Solving courts or these domestic violence courts. And that's that's what we study, so many argue that the courts have dismissed domestic violence as a private family matter and I would argue that it's definitely not a private family matter, although of course, it occurs within families and the criminal justice system asks people to talk about very personal things in public, and these are often very painful.
Kerri [00:05:34] So what domestic violence, courts do as they bring together domestic violence professionals. Most central to this paper is going to be the judge and the court staff who have some training in sensitivity and specialization to sort of why this is hard, why victims may have a hard time testifying or following through with charges, what dynamics to expect and in particular, they may have some tools to overcome those challenges. So in like the mid-1990s, we started to see the emergence of two basic kinds of domestic violence courts. The first kind domestic violence problem-solving courts is the version most commonly evaluated in the literature and I should say early on, this is actually not the kind of version of domestic violence courts that we investigate problem-solving courts like in their purest form in the way they're typically seen in the literature arer courts that allow defendants to opt in to a diversionary court system and that system runs parallel to the traditional system. It's premised more on like therapeutic jurisprudence.
Kerri [00:06:35] And there's a lot of variation, but defendants have judicial monitoring, better intervention programs and other social services. And should they misstep either by not complying or re-offending then judges may have the authority to hold them in contempt, implement a criminal sentence or even just like refer them back to the traditional court system. The second kind of domestic violence court is a specialized domestic violence court and that's the kind that we're studying in this paper, and it's sort of lesser known in the literature lesser sort of dealt with in the literature. And this is one where domestic violence is treated explicitly as a crime. It is not diverted from prosecution. The goal is not necessarily therapeutic jurisprudence, especially for the defendant. The goal is to adjudicate the domestic violence case with the standard tools available to the court, but to do so with a dedicated staff, which in our paper is going to include specialized judges, those judges' probation officers and judges' court officers.
Kerri [00:07:38] So those are kind of the problem-solving courts and specialized courts, in a nutshell.
Jennifer [00:07:42] So before this paper sounds like these have been around a while, but before this paper, would it be known about the effects of these domestic violence courts on various outcomes we might care about?
Kerri [00:07:52] Yeah. So we knew sort of the most about that first kind of court, that problem solving court that I described in a great sort of starting place for that is a 2016 meta analysis done by Gutierrez and colleagues and they looked at 20 peer reviewed and non-peer reviewed papers and these papers primarily, like I said, focus on that first kind of domestic violence court, that therapeutic problem-solving court. And they concluded that, yeah, on average, most studies found that defendants whose cases were handled in this DV court are less likely to recidivate, but when they sort of accounted for credible research designs and sort of put more stock there, maybe research designs that use pre or post-match comparison groups, they tended to find really small or null effects of these problem-solving courts. So we're excited to study the second kind of specialized domestic violence court, which I haven't really seen in the literature, so I don't think we know that much about this particular court setting. And so this allows us to really build on the research and build on our knowledge about domestic violence court interventions and like this contribution. Also, I should note, we're going to look at this court sort of several years after it's been up and running. Most evaluations look at a court when its new when it's first happened, especially if it's using a pre post design. So what we think is really novel about this paper is the setting in Davidson County or Nashville, Tennessee, is a setting that allows us to build on the research and specifically look at these specialized domestic violence courts, which is that second kind of domestic violence court that I mentioned, it's the one that operates with specialized judges and in the more standard criminal justice system.
Kerri [00:09:40] So in addition to looking at the second kind of court, which doesn't seem to be really present in the literature, at least not to our knowledge, we're going to look at and evaluate a court system several years after it's been up and running. So many of the other studies, especially those with a pre post design, are looking at case outcomes resulting from a new intervention. And so finally, and probably the most important, we're going to use an identification strategy that closely mirrors random assignment, and so hopefully we're generating credibly causal effects of these courts.
Jennifer [00:10:14] Yeah, and so what makes this question so difficult to answer? It sounds like you had real world experience working in these courts so presumably, you started your research career wanting to answer questions exactly like this one. So what were the challenges that you had to overcome as a researcher? Is it mostly data availability or mostly finding those good natural experiments? Or is it both of those things?
Kerri [00:10:38] It's definitely both of those things, but I think it's the natural experiment is particularly, you know, a challenge in the setting like as one of my grad school buddies, Cody Wing used to say, and I don't know, he probably still does counterfactuals are just really hard, right. And often it's really hard to get over endogenous case assignment, especially after a judicial reform. So after there's a domestic violence court, then if all of the domestic violence cases go to the newly formed domestic violence court, researchers can absolutely employ a pre/post design to try to understand the effect. But then there's just not a clear counterfactual to account for secular trends or shifts in judicial shocks, right? And also, like, as I alluded to earlier, the pre/post design really limits the evaluation to studying the effects of the domestic violence court at or around implementation.
Kerri [00:11:33] And that's just probably not a great measure of domestic violence courts overall or like in the long term. So like, for example, I think it's probably going to take some time for attorneys to understand judges preferences or their reasoning. And it may take time for the judge to see repeat victims and defendants or even get there like sentencing speeches down, which which I think is actually really important role that judges play. I'm convinced that courtrooms have a personality and that these things just need a little time to settle in before a researcher can or maybe should measure the long term effect of a domestic violence court, or probably any intervention. But like outside of research design, we definitely had some struggles obtaining the courtroom data like, even though I intimately knew all of the data that was available, it was still really hard and a challenge, but you know, Aria are amazing coauthor and current job market candidate wrote the code to web scrape the data and then we got the data and it was it was fantastic. But even then, like there wasn't a code book per se and so we used my institutional knowledge and my memory and some phone calls to lots of colleagues that are still there to answer our questions about the data and make sure that we knew everything that we had. But like this is tricky, and I think it just speaks to the importance of researchers to spend time learning the inner workings of these kind of evaluations or to team up with someone who knows all the little quirks of the of a system or an intervention that they're trying to estimate.
Jennifer [00:13:05] Yes, it definitely seems like an ideal situation. Someone like you on this research team.
Kerri [00:13:10] And it was an ideal situation for me to work with them, I hope.
Jennifer [00:13:16] So in this paper, you're focusing on domestic violence courts in Davidson County, Tennessee, where Nashville is located, as you mentioned. So give us a bit more background on how these particular courts work and to start out how are judges selected to serve on the courts in that county.
Kerri [00:13:31] Yeah, great. So starting with the judges in the court that we're going to examine that we're going to use the mist, which is going to be the misdemeanor jail docket, that docket, it's part of the General Sessions Court of Davidson County. General Sessions Court is characterized as like a high volume court, they hear all kinds of cases civil cases, criminal traffic, environmental metro ordinance violations. But there are 11 judges that serve in general sessions and each judge is elected to their own eight year term, and each judge really presides over a division, which is like a concept we'll come back to. But the general sessions judges have a lot of autonomy over how dockets work and actually, in the mid 1990s, one of the general Sessions judges, Judge Faymann, was particularly moved by a domestic violence case, and Judge Faymann came away from this case, really believing that court specialization was something that Nashville would benefit from and it would improve domestic violence outcomes. I'm sure there were tons of like back room conversations about this, but from there, the general Sessions judges decided to have two, two of the 11, general sessions judges and divisions dedicated to hearing domestic violence bond docket cases to the bond market is different than the jail docket. I'm going to promise to cover all this, but I just want to highlight that. So those very first domestic violence judges were Judge Kenny Harrington, who has since retired and Judge Gail Robinson, who actually still serves on the domestic violence docket today.
Kerri [00:15:07] And so starting in the mid 1990s, these two judges were the only judges to hear domestic violence cases on the bond market, and I should clarify. Bond and bail are used synonymously in this context. So the DV bond docket exclusively hears cases with defendants who posted bail or bond, whichever your preferred term is and those defendants are charged with a domestic violence related crime. So each judge is really the head of a division. All of the judges have equal authority, but each judge has two dedicated court officers and two dedicated probation officers. And these court officers and their probation officers move with a judge like they're a team, right? And so the bond docket that was created in the mid 1990s has more than just specialized judicial divisions or judges.
Kerri [00:15:58] The district attorney's office in Davidson County has a family protection unit, so they had DAs that were dedicated to prosecuting domestic violence crimes as well as victim witness coordinator, that's what I was, I was a victim witness coordinator that were dedicated to assisting and DV prosecution. Nashville has batters intervention programs, they have a rich community of domestic violence advocates, which has kind of grown over the years, and the public defender's office also made pretty regular assignments to the DV bond docket. But of course, you know, defendants could also use private counsel if that's something that they wanted to do. So I know that's a lot of information, but like the takeaway here is that after 1994, if a domestic violence defendant posted bond, their case was going to be heard by a specialized domestic violence judge who had dedicated probation officers, dedicated court officers. The case would be tried by dedicated D.As.
Kerri [00:16:52] There'd be a robust network of victim support professionals. And along with some really good options for the defendant spatters intervention programs. And it's possible, you would say to me, why aren't we evaluating the bond docket? That sounds like a great docket to evaluate. But the problem is there's just not a good way to do it. So court records did not become really electronic until the year 2000, so we can't employ a pre post design. It's just not possible here, right as it started in 1994. We're not going to get electronic documents until 2000, and all the domestic violence misdemeanor that post bond have their case set here. So like I said before, there's no counterfactual for the bond docket. It's just really hard to evaluate the causal effect of the bond docket for domestic violence, the stimulants, even if that is a question that we're interested in.
Jennifer [00:17:45] Mm hmm. But luckily, you also have this jail docket. So as you've mentioned from these misdemeanor domestic violence cases, the way the case they're assigned to courtrooms is going to depend is going to differ a little bit for people who are able to make bail or bond or whether they're not able to make bond. So tell us about the process now for the jail docket.
Kerri [00:18:07] Yes. OK. The jail docket is going to introduce some really important variation for us. So maybe even just backing up a tad more like at the crime scene, the police officer in Tennessee is encouraged to make an arrest. They have a kind of pro arrest law and they make it arrests. If there's probable cause to believe that a probable cause on probable causation, that's amazing to believe that a crime has occurred, right? And then the officer is going to take the defendant down to night court, which actually operates at all hours of the day, not just tonight.
Kerri [00:18:41] And their commissioner is going to review the charges and issue a warrant. The commissioner is going to set a bond. And while we don't have bond amounts and our data, my memory is that bonds were typically about one thousand to five thousand dollars per warrant. So in Nashville, if the defendant can post about 10 percent of that working with a bail bondsman plus some administrative costs, they can typically post their bond right? And about half of domestic violence misdemeanors in Nashville in our sample post bond. And for the other half, they don't post bond at their cases, go to the chip docket. So we are going to hone in on these other half of cases that are not posting bond.
Kerri [00:19:24] Their cases, like I said, are going to go to the misdemeanor jail docket. So there's an institutional feature that turns out to be really important for us and that is that the day of the first setting on the misdemeanor jail docket is 100 percent dependent on the time of day and the day of arrest. Right. So if a person is arrested on a Friday between 4:01 a.m. and Saturday at 4:00 a.m., their case will be set on that first Wednesday that follows. It does not matter how long the Wednesday jail docket is. That's where the case goes. And yes, if you're thinking that that Wednesday and Thursday, jail dockets are super long. They are brutally, brutally long. It does not matter. That's where the case goes. So defendants have this like predetermined scheme for when their case is going to be set. And importantly, so do judges. So the general sessions judges, their dockets are like booked out. I don't even know for how long, like I think for like, maybe forever like they're just booked out.
Kerri [00:20:29] Like, you can go look right now and know who's going to be on the jail docket four months from now. It's just a schedule that's set and they just rotate, right? And so what's nice for us here is the idea of like judge shopping or trying to select a judge for a particular case is really just like not feasible in this institutional setting. And of course, like, you know, we're going to formally test for this in a balancing test to make sure that the defendants that go before domestic violence divisions look like the defendants that go before non-domestic violence divisions. But like, that's what we find, and it's because of the institution's arbitrary nature in assigning defendants to judges. It's just very arbitrary.
Kerri [00:21:13] So the defendants are set and the judges in their division staff are just like rotating through the misdemeanor jail docket, right for a week. One Division one is there, week two, Division Two is there and so forth. And then they just repeat. One more thing that that is really important, though, about the misdemeanor jail docket is those specialized DAs and victim witness coordinators that are a feature of the domestic violence bond docket. They're also assigned to work the jail docket. So importantly, the only thing that's like rotating and varying and our context are the judicial divisions, right? And so this is another contribution of our paper is it actually allows us to estimate the effects of judicial specialization as opposed to like the entire courtroom and so we kind of think that we're unpacking the black box of these courtrooms a little bit. We're really able to hone in the effect of a specialized judge as opposed to just like a specialized courtroom setting all misdemeanor cases where defendants cannot post bond go here.
Kerri [00:22:16] So it's a misdemeanor jail docket and it is a mix of domestic violence and non domestic violence cases. And so we're actually going to use this in our specifications to tease out the judge specific effect relative to the domestic violence specialization. Specific effects, right, so we can see how a judge rules in both domestic cases and non domestic cases that are set on the same day.
Jennifer [00:22:41] Right. OK, so let's see. So you've got someone who's arrested on a Friday for domestic violence this week, and they get assigned to a particular courtroom on that following Wednesday, and then someone else is arrested for the exact same charge the following Friday. Their case is going to be heard that following Wednesday, and you're going to have just sort of by chance, you're going to have different courtrooms that you wind up in as these two defendants and one of them might happen to be a domestic violence court just by the luck of the draw and this rotation schedule and the other one is not. And so that's kind of the core of your your natural experiment here. Am I getting that right?
Kerri [00:23:19] You are right. So basically, each defendant has about like a two in eleven or 17 percent chance of being before a domestic violence general sessions judge and in fact, that's what we find about 17 percent of the cases go before these domestic violence judges sort of reinforcing that. You know, it's just all happening by chance.
Jennifer [00:23:41] Yeah. And I also want to highlight, I think you said that about 50 percent of these misdemeanor domestic violence defendants are posting bond.
Kerri [00:23:48] Yeah, 50 percent are posting bond and about 50 percent are not posting bond. So it's like a half and half split.
Kerri [00:23:53] So yeah, it's not a small part. It's definitely, you know, half and half.
Jennifer [00:23:57] Yeah, that's something people might have been worried about otherwise, that this is sort of like a weird sample, but half half is not weird.
Kerri [00:24:04] No half is not. We are definitely a little bit different. They're certainly more indigent. For example, they can't post bond. They are a little bit different, but it is half and half.
Jennifer [00:24:13] OK, great. So you're going to use this as if random assignment of cases on the jail docket across these domestic violence and non domestic violence courtrooms as a natural experiment, as I mentioned. So how exactly do you use this to measure the causal effect of assigning a case to a court?
Kerri [00:24:31] Sure. OK. So we have a setting where misdemeanor defendants, some of which have DV case and some don't, are getting set on the same exact jail docket at the same exact time. And like I said, we have a setting where some weeks two out of eleven, those defendants are getting their case before a TV division and some weeks about nine out of eleven defendants have their case set before a non DV division. And again, the only thing that predicts whether or not the defendant has their case in front of the TV division is the day and the time of their arrest. Nothing about their characteristic predict this, right. And so once we can establish that which we do, we know do the balancing test and compare the, you know, attributes of the defendants between the two settings. We can just use OLS. You can just use we can just use regression. We're off to the races, which is amazing. We run our specifications also like with and without controls and consistent with this idea of random assignment, the controls don't really do anything to our DV court estimates.
Kerri [00:25:33] They just make our standard error smaller, which is, you know what you learn a metrics 101, and it's it's just great to see it happen in practice. So we do this in like two ways. We start by just estimating a difference model. So for any given outcome, we'll talk about those morre soon about like for any given outcome, let's just say the defendant is guilty or not guilty. We can regress that on an indicator variable equal to one if the case is heard in a DV version and zero, it's not and like, of course, time effects. We do it both ways, like I said, with or without control. They don't seem to matter. But the problem with just the simple difference, if you will, is that we worry some judges might just be more lenient or strict compared to others. Right. And there's like a literature about this, right? Like people were, you know, think about exploiting different kinds of judge effects.
Kerri [00:26:24] So what we can do and what we do to account for that is to do the difference in difference and when we are able to do the difference in difference this allows us to understand how domestic violence judges are making decisions differently. But the problem with that just straight difference specification is that we worry some judges might just be more lenient or strict compared to other judges, and there's a literature that's pretty consistent with that. So we can and do actually account for another difference. So like I said, on the misdemeanor jail docket, all of the cases DV and non-DV get assigned to whomever is just on the bench that day. So when it's a non DV division, we say hear DV cases, they hear, not DV cases. When its a DV division ewek division, we they also hear the same thing.
Kerri [00:27:12] So we use this to create our preferred specification, which is just a really straightforward difference in difference. And this allows us to understand one, if DV judges are making decisions differently than how they themselves might rule on other non DV cases. And two, if they're making different rulings in DV cases relative to their non DV tier judges.
Jennifer [00:27:35] Right. So if the concern here is basically like, maybe this just a special kind of judge, maybe a more lenient judge tends to be the one that wants to be a DV in the DV court. And so any case then that goes before that judge might, might have a lower conviction rates or something like that. And so we're just looking at the domestic violence case. So as you might attribute that to the fact that it's a domestic violence court versus the selection of the judges. And so that's what that difference in difference helps you resolve.
Kerri [00:28:01] Yes.
Jennifer [00:28:02] So tell us more about this amazing data that you scraped. What data do you use for all this?
Kerri [00:28:07] Yeah. So our data, I think, comes from about four sources. So, you know, we're first going to look at what happens in court and our our analytic data from about 2000-2006. And I should say our analytic data is only going to have male defendants for those day of court outcomes. And for that data, we use the Davidson County Criminal Clerk data, which Aria obtained through web scraping and he did that all the way up through 2018 so we can look at future crimes as well. And this includes all the case information, all the courtroom information, like when and where a case is that what the defendant is charged with and how the case gets adjudicated and the defendant characteristics from the warrant.
Kerri [00:28:48] And then, of course, we're going to control for judge characteristics like the judge's sex and I knew that although that was also really that was easy to get and how many years they served on the bench. We just like looked that up. And then after we establish if domestic violence and non domestic violence divisions have different court outcomes, then we're going to also look to see whether or not the defendant re-offend if they show back up in court later, right? And so each defendant has what's known as a unique OCA numbe, it's like their Social Security number for people in jail. And so it's really easy to sort of trace defendants into the future and so we can see if they show back up in court again three years later, using that same web scraped data.
Kerri [00:29:35] And then the Metro Nashville Police Department also gave us their police report data, and this data includes all police report, some of which are going to materialize into warrants and some won't. And we're going to use that data to in particular to see a victims call the police in the future. And if those police reports are also going to tell us if the victim cooperates with the police in the future. So we're going to use that to look at future victimization as well.
Jennifer [00:30:03] OK, and what outcome measures are you focused on?
Kerri [00:30:06] Yeah. So we are focused on a lot I think about. But the first thing, like the first order question is like what happens to these cases in court right. So are defendants more or less likely to be convicted if they're assigned to a DV division conditional on conviction? Like, are they more likely to be incarcerated? And then the maximum sentence length and just kind of like give a teaser to say we had some results that believe really shocked me and like totally did, but with my priors.
Kerri [00:30:37] And then, of course, I guess I sort of previewed, we're going to ask if assignment to DV divisions reduces defendants future court appearances or future offending and or changes the victim's future victimization, which we measure by whether or not the victim calls the police, files a warrant and if the victim refused to cooperate with the police at the crime scene.
Jennifer [00:30:55] Results that don't fit with your priors are the most fun kind of results.
Kerri [00:31:01] They are kind of once you figure them out for sure, but definitely some head scratching. In the meanwhile, yes.
Jennifer [00:31:09] All right. Well, let's talk about those results. So what do you find is the causal effect of assigning a DV case to a DV court on the case outcomes?
Kerri [00:31:19] So we're going to look at two kinds of cases. The first is among, you know, DV cases, which we defined as any case that includes a charge for a domestic assault and these are usually class-A misdemeanors. And what we find is that the probability of conviction is actually about eight percentage points lower when cases are assigned to a DV division relative to a non-DV division. I did not expect that this estimate is like barely really consistent in both our difference and our difference in difference strategy. So that tells us that all of the judges are convicting non DV crimes with similar probabilities. But the specialized judges are just much less likely to convict in particular cases, so that that really shocked me. And when we presented it to Nashville domestic violence professionals, they were pretty shocked by that finding tip.
Jennifer [00:32:15] Yeah, because on its face, it sounds like the judges are going easier on the DV cases.
Kerri [00:32:22] That's kind of what it sounds like on its face, right? Definitely I did not fit with my priors. It didn't. It didn't fit. But look, we're into that. Don't worry.
Jennifer [00:32:31] OK, all right. We'll talk more about that momentarily. OK. So then tell us what you find for other case outcomes.
Kerri [00:32:38] Yes. So the next thing we look at is incarceration conditional on conviction, right? And here and our difference model, domestic violence judges seem about five percentage points more likely to incarcerate, but I will say like in our difference in difference model, the effect remains positive, but it shrinks and isn't significant. So like, what we take from this is that while assign that to a DV division is more likely to result in jail time if the defendant is convicted, that finding is not unique to domestic violence defendant. So in that sense, domestic violence charges are a little bit stricter. They're sort of more likely to incarcerate anyone conditional on conviction.
Jennifer [00:33:19] OK. And next, you're able to look at the effects of DV courts on recidivism. So what do you find there?
Kerri [00:33:24] So we look at if the offenders show back in court within three years. Right. And so I should just say these are again misdemeanor domestic violence cases and if a person is sentenced to jail, most of those jail sentences would be like 11 months and twenty nine days at 30 percent, which is like eighty one days on average, right? So we don't have large incapacitation effects. This is the point of that. So and we look at showing back up in court and a lot of ways we look at like, did you come back for domestic violence assault charge? It could be a felony or misdemeanor. Do you come back for any assault charge, also a felony or misdemeanor? Do you come back for like any criminal charge at all? And then we just specifically look to see if you came back for a domestic violence felony charge.
Kerri [00:34:09] If we really wanted to look at that last outcome in particular because we think it's a proxy for escalating violence and we would just really want, you know, we were really interested to know if that was affected by the domestic violence division. But the answer is no. Domestic violence divisions don't seem to affect defendant recidivism. It doesn't matter if the defendant is assigned to a DV or non DV judge, they're just as likely to show up in court and the next three years, basically like no matter how we defined it. And this is somewhat surprising again, given that defendants were less likely to be convicted in the domestic violence court.
Jennifer [00:34:46] Mm hmm. So I think many people might be surprised to hear that domestic violence is often a misdemeanor.
Kerri [00:34:52] Mm hmm.
Jennifer [00:34:53] So do you just off the top of your head? Like what share of domestic violence charges are or misdemeanor versus felony in this court?
Kerri [00:35:00] So I actually used to do the statistics for the DA's office while I was there.
Jennifer [00:35:07] What a coincidence.
Kerri [00:35:09] And I would say that probably about twenty five percent or so cases were felony domestic violence cases, and those cases are actually heard on the felony jail docket. So there's a whole nother jail docket for felony cases. We don't include those because there's not a lot of variation in the outcome. Most of those cases are just bound over criminal court, so they're sort of knocked upstairs. General Sessions doesn't have jurisdiction to adjudicate felony cases. So we don't look at them because there's not a lot of variation in the outcome, but about twenty five or 30 percent of cases in Nashville are felony domestic violence cases.
Jennifer [00:35:46] OK, so most of the domestic violence cases are misdemeanors.
Kerri [00:35:49] Absolutely, yeah. All right.
Jennifer [00:35:51] So let's talk about a third set of results. So you're able to look at the effects of domestic violence courts on re victimization. So what do you find there?
Kerri [00:35:58] Yes. So this analysis as a former victim advocate was super important to me, right? So we really wanted to know if the victim in the initial case contacted the police at a future date. And I should say here, we're only able to use our difference model. The difference in difference model just doesn't really apply in this setting because non- victim cases, there's no comparison group. So we're only going to use our difference model here. And we look at whether or not the victim called the police for any incident within three years of the initial case filing and then whether or not the victim filed a police report for assault specifically. And basically, we find that victims whose defendants had their cases set in a domestic violence division are about eight percentage points, which is like 16 to 20 percent less likely to call the police in the years following relative to victims who had their cases and non DV divisions. This finding by itself is is like a little ambiguous. So are victims are not calling because they're safer? That would be good. Or is it because they had a bad court experience and don't want to reengage? That's you know what the literature refers to as a gag effects only one of those is good and and the other is pretty clearly undesirable. So next, we look at whether the initial assignment to a DV division affects the probability of being a victim in a warrant conditional on calling the police and those coefficients are all positive, but they're not significant.
Kerri [00:37:28] So finally, our last measure uses a particular field in the police report and not feel this checked if the victim refused to cooperate with the police at the crime scene. And what we see is that victims whose initial case is set before a DV court are less likely to refuse to cooperate by about five to seven percentage points, which is like forty two to 62 percent. It's really big. So if you say that in the positive like victims whose cases were assigned to a domestic violence division are more likely to cooperate with the police at future crime scenes, of course, I think still conditional on calling the police. We think it's critical to think about all of this together. Like, like no one piece, I think paints a story. But if you think about it all together, I think you get a consistent story that's important, right?
Kerri [00:38:17] So even though defendants whose cases are set in domestic violence courts are less likely to be convicted, they're just as likely to return to a docket. Importantly, we don't see any evidence of escalation and their criminal charges, i.e. like a transition to felony charges and the relevant victims call police less. But when they do, they're much more likely to cooperate with the police. And it really seems to us that domestic violence divisions are moving us towards to something that's a little bit more socially optimal, right? They seem to be reducing like type one error and misdemeanor conviction by declining to convict in some cases. But they're doing so without increasing type two error, which would happen if we can if we did not convict someone who was dangerous. Right. So they're being more discerning with who they convict in a way that's consistent with victim safety.
Jennifer [00:39:12] Yeah, it's super interesting. So and then you do a whole bunch of other stuff to dig into the data a little bit more, both to consider the robustness of your main estimates and other potential mechanisms. So tell us about some of those checks and tests that you run. What do they tell you?
Kerri [00:39:27] So they inspire some confidence.
Jennifer [00:39:30] Always what you want.
Kerri [00:39:32] So good. So, you know, for robustness, we do like a whole bunch of things with the outcome. Like, should we model this at the case level, the charge level, should we include all the victim level misdemeanor crimes like harassment, stalking, right, our main things is just sort of consider assault charges and domestic violence assault charges. And no matter how we construct the domestic violence outcome variable, the results hold. And that's what I mean by like inspired some confidence. But of course, like our first finding is still troubling, right, that the probability of conviction is lower in domestic violence division is still puzzling. And in order for that result, to be credible to us and to others, we really needed to know why. And so we dug in. So a case has two basic outcomes. It can be dismissed or the judge can find the defendant guilty. Right. Dismissals often happen because the victim doesn't want to cooperate. But certainly other reasons too, like maybe the warrant just didn't actually allege a crime, right? So what's interesting to us is dismissals like that are really evenly distributed across the Adnan division. So like that is not driving our findings. So then we ask, well, how does one get a guilty verdict? So either the defendant pleads guilty or there's a bench trial and the judge makes the determination. Remember, we don't have jury trials, so this is all determined by the judge.
Kerri [00:40:54] So what's actually driving our findings is that in a domestic violence division, those cases are just much more likely to go to a bench trial. And whenever you have a bench trial that introduces uncertainty, right, a plea like is by construction going to have a 100 percent conviction rate. But a bench trial like even with the best case, like there's there's uncertainty the judge just may not agree to convict for a host of reasons and that is what's driving the lower conviction rate. The cases are just going to trial more, which is not necessarily a bad thing. Right. So cases might go to trial more if the D.A. feels more confident in their ability to convict, or if the D.A. is pursuing a stricter sentence and the defendant is unwilling to plead, or if the judge refuses to take a plea agreement because, you know, they think it's it's not strict enough. So like the latter scenario is a little bit more subtle, but for example, Judge Robinson who is a DV judge would often like, want to see compounding plea agreement.
Kerri [00:41:58] So if the defendant had been found guilty in the past, he would want that new sentence to be stricter. And everyone just kind of knows this in the courtroom like this is by evaluating at the pre post time period may not be the best like, this definitely takes some time for everyone to sort of figure out and to know your judge. But it's going take to kick things into a trial in sort of an implicit way. And our data, just unfortunately like this, is the drawback of the story of data. Our data doesn't allow us to know if trials are happening more because judges want to be heard more often, which would be consistent with like Rich Lynskey and colleagues and their 2006 paper, or if DA's, are just more willing to take risks.
Kerri [00:42:40] Like my feeling from working in these courts for several years is that it's like a little bit of those like it's a little bit the D.A. It's a little bit the judge. And unfortunately, we don't know which one in which case to sort of figure out which one dominates.
Jennifer [00:42:52] But overall, it just basically makes all of it makes that surprising results make more sense.
Kerri [00:42:58] It does make that surprising result, make much more sense. And again, like it's not necessarily bad that cases are going to trial and that victims are being heard or that defendants are are having their right to trial. That's not necessarily a bad thing.
Jennifer [00:43:12] Right. And if the marginal cases that are now not winding up with the conviction, it's not driving the recidivism rate up, then that seems like a good thing.
Kerri [00:43:21] Right? It's not driving the recidivism rate up and victims citizens wishing to engage with the court system.
Jennifer [00:43:27] Yeah. All right. So what are the policy implications of these results? What should policymakers and practitioners who are listening take away from all this?
Kerri [00:43:36] So I hope a lot. So like I think this is the domestic violence specialization seems important, right. Conviction seems more targeted without any tradeoffs. And we do think that this also has just some really practical implications. So the work implies that if the jurisdiction has a domestic violence court, then everyone in that courtroom should expect more trials. And this is important for people to be prepared for trials, especially if you want to keep victims reengaged with the court system if they need it right. So prosecutors, these are misdemeanor crimes. So these are likely to be prosecutors that are early in their career need to have trial preparation, victim services need to be preparing victims.
Kerri [00:44:18] Officers may be called to testify, which can drive up some court costs because you pay those officers to be there, but everybody seems to be prepared for more trials. I think as is probably one of the key findings of this paper.
Jennifer [00:44:31] What other papers related to this topic have come out since you all first started working on this study?
Kerri [00:44:36] So there have been there are some exciting work. So Ashna Arora and her colleagues study the impact of specialized prosecution on the safety of domestic violence victims, and they look at this in Cook County, Chicago. And so we're looking at judges, they're looking at prosecutors, and I think this is just a nice compliment. And their results indicate that specialized prosecution, which increases prosecutor capacity on each case, introduces victim advocates into the prosecution process, et cetera that lowers the likelihood of homicide for those on the margin of of inclusion, of getting this increased prosecutor capacity or not. Right. And so this is this is just sort of saying that it's not necessarily one thing that we should do. It's not necessarily just like domestic violence judges. It's it's sort of really a system that we need to think about in each part of the system has its piece to play. And then of course, there's there's work by Amanda Agan, Jennifer Doleac and Anna Harvey about misdemeanor prosecution and I think you correct me if I'm wrong, but that paper finds that non prosecution of nonviolent misdemeanors leads to a large reduction in the likelihood of new criminal complaints, and that those effects are greatest among first time defendants. So like obviously there a difference here, right. Domestic violence is a violent crime, and all of the cases in our sample are prosecuted.
Kerri [00:45:58] But our work also indicates that the criminal justice system can do better right in terms of figuring out convictions and specialist DV judges may be one way to do that right to better target conviction so that we're doing that in an optimal way.
Jennifer [00:46:15] Yeah, that's a great point. It's interesting as our our research team talks with prosecutors offices the year after we put that paper out, it comes up again and again that a lot of offices are trying to figure out how to handle these misdemeanor domestic violence cases, like where to draw the line, who should be prosecuted, which ones should we move forward with. So, yeah, lots of the I agree it's an interesting intersection there. So speaking of additional questions that need to be answered, what's the research frontier? What do you think the next big questions are in this area that you and others will be thinking about in the years ahead?
Kerri [00:46:46] Oh my gosh, we ever stop thinking so.
Jennifer [00:46:51] Endless questions.
Kerri [00:46:53] So, you know, as like a former victim advocate, like, I have a lot of questions about what works and what doesn't and how we can better serve communities. But so as we talked about, like this particular paper is focused on a very specific population, and all of these defendants are most likely indigent because bonds are fairly low and they're not able to post them. And also, the defendants in our sample have a lot of contact with the criminal justice system. And like we said, this only describes half the defendants in the court system, right. This misdemeanor defendant. So I still think this question that's out there in the literature like, you know, what might the causal effect of specialized courts for other defendants like, namely those who can't post bail or bond be like that question is still looming.
Kerri [00:47:38] The literature has grappled with it, but I don't think it's resolved. And then, of course, like what's the causal effect of other interventions, like the work that Ashna and her colleagues are doing I think time to trial is another thing that we really need to think about, right. So how long a case where defendants can post bond, how long between when the offense happens and when when that case is heard before a judge, it's just really important. And I don't think we understand how important it is and I don't know maybe my priors is wrong, but my prior tells me it's a very important thing to think about.
Kerri [00:48:14] And then I think like the other thing we just really need to think about is scalability of services. It actually took Nashville a little over a decade to create a system where all DV cases the jail docket and the bond docket could be served by a DV division judge. So like our paper stops in 2006 because in 2006, that's when all of the jail docket cases were also heard by a judge so it took a like a full decade to get this program up to scale. So like, how do we find good interventions and then implement them or figure out the parts that are most needed and implemented? And you know, I think the Yotam Shem-Tov, Steven Raphael, and Alissa Skog paper about like restorative justice and understanding that, you know, things like that make it right program have effects. But like understanding like scalability is just going to be something that we're going to have to continue to struggle with in resource scarce environments.
Jennifer [00:49:11] My guest today has been Kerri Raissian from the University of Connecticut. Kerri, thanks so much for talking with me.
Kerri [00:49:17] Thank you.
Jennifer [00:49:24] You can find links to all the research we discuss 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 Emerson 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 Hayley 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.