Stitching Together America’s Criminal Justice Data with Mike Mueller-Smith

The criminal justice system is big, it is complicated, and it is expensive. But how do we know what parts of it, if any, are working? 

Enter Mike Mueller-Smith.

A decade ago, Mike started building the Criminal Justice Administrative Records System (CJARS) a data infrastructure project that links criminal justice records across 40 states to employment outcomes, health data, and more. It is the kind of tool that makes you wonder why it didn’t exist sooner.

In this episode, we get into why criminal justice data is so fragmented, what it takes to stitch it together, and what researchers keep finding once they can actually see the full picture. We also talk about Mike’s Justice Outcome Explorer, one of my favorite criminal justice data sites out there – a public dashboard built to help average Joes make sense of complex and sensitive criminal justice data.

Mike Mueller-Smith is the Co-Founder and Director of the Criminal Justice Administrative Records System (CJARS for short), a data infrastructure platform that is modernizing American research and statistical reporting capacity. He also is an Associate Professor of Economics at the University of Michigan and does research on the multi-faceted determinants of crime in the U.S. and the broad-ranging consequences of our criminal justice policy.

Transcript

SPEAKER_00 0:00

Well, I'm Jeff Asher, and this is the Jeffalytics Podcast. The United States spends hundreds of billions of dollars a year on its criminal justice system and has almost no unified way to measure what any of it is actually doing to the people who move through it. Mike Mueller-Smith is an economist at the University of Michigan who decided to do something about that. A decade ago, he started building the Criminal Justice Administrative Records System, or CJRS, a data infrastructure project that links criminal justice records across 40 states to employment outcomes, health data, and more. Is the kind of tool that makes you wonder why it didn't exist sooner. In this episode, we get into why criminal justice data is so fragmented, what it takes to stitch it together, and what researchers keep finding once they actually see the full picture. We also talk about Mike's Justice Outcome Explorer, a public dashboard built to help the average Joes make sense of complex and sensitive criminal justice data. Let's dive in. My guest today is Mike Mueller-Smith. Mike, thanks so much for joining the program. Thanks for having me. Glad to be here. So, Mike, take me through what is your background? What brings you here today?

SPEAKER_01 1:06

So I am an economist at the University of Michigan, and I do research on crime and justice in the United States, thinking broadly, like why do people engage in illicit activity? What are the consequences of the policies that we have to try to control crime? And I started a data infrastructure platform about a decade ago called the Criminal Justice Administrative Record System, or CJARS for short, which helps me do my research and is basically an effort to try to modernize our capacity to understand the state of our justice system, how what happens to people after they go, they go through it, and a whole range of things.

SPEAKER_00 1:46

So let's start with the easiest question first. Why do people engage in criminal activity?

SPEAKER_01 1:52

I mean, that's a complex question. I, you know, I it's it's easy to say, it's hard to answer. I think if we had the answer to that, you know, people like me might be out of a job. You know, I think there are a whole range of things that people engage in crime. And I think the hard thing about answering that is that crime is kind of a catch-all term for a lot of different types of unproductive or antisocial behavior that can cause harms in society. And so you have a range of things from violence to property crimes to drug use. And so the motivations also are going to be quite varied. It can either be unmet financial need, unmet health needs, a potential lack of deterrence to get people to follow laws and kind of falling through on the punishments that they might face. And so I think that's actually a pretty complex question to try to answer.

SPEAKER_00 2:44

I guess more broadly, why would you as an economist study crime? Why is that something that economists study so much of? And what does the field of economics bring to this field of research of this area that sort of other, you know, your criminologists necessarily might not necessarily get?

SPEAKER_01 3:03

There are a ton of reasons. So I talk about this, I teach undergrad and grad courses on the economics of crime. And so I always start out with motivating. Like, why is it that I, as an economist, care about these issues? And you can come at this from a variety of perspectives. One is just that we spend a ton of money between the federal, state, and local level on the justice system. You know, you think about how much we're spending on policing, corrections, courts, the whole gamut. That's going to be in the hundreds of billions of dollars per year. At the same time, what we do to people who go through the justice system, things like permanent criminal records, interventions like incarceration, as well as more positive things like rehabilitative programs, these have the potential to reshape people's lives, either for better or for worse. And then as a last thing, obviously, the biggest motivation as we think about crime control policies are the what we call in economics externalities that are imposed on the general population, the harms that people experience as a result of someone deciding to engage in crime, whether that's you know, somebody engaging in violent conduct or property crimes, that's going to leave somebody as a victim and create uh social harms. And so people have tried to estimate what the total costs in society are. And those some estimates suggest up to trillions of dollars. So there are a variety of different reasons why I think economists come to approach this topic, whether it's from thinking about the efficient use of public funds, the effectiveness of criminal justice policies on those who are being directly as well as indirectly impacted by them, you know, think about household spillovers, or in terms of kind of controlling the scope of victimization and the potential harms that can come from that. And so I think, you know, as an economist, uh, what we think about are kind of disciplined empirical research that really puts causal designs at the forefront, deep experience with kind of large administrative data sets and kind of trying to push the frontier in terms of what we can measure, as well as an appreciation for kind of the um the trade-offs that we face in society. So, you know, if we want to amplify something like deterrence, that might come at the consequence of creating collateral impacts on individuals and communities. Conversely, if we want to completely pull back and take something like progressive approaches to the justice system, that's going to have its own potential negative ramifications. And it's really about trying to figure out what the right balance is for all of it. You know, how do we best design something that makes the most people better off?

SPEAKER_00 5:50

So, what is C Jars? You mentioned it, criminal justice administrative record system. You know, why did you build this? If are you like me that when you you find a need that, like I wish I had X to do my job better, and then you're like, well, why don't I just go build X? Is that sort of the why it came about? Or what what did this come from and why do you think it's so necessary?

SPEAKER_01 6:12

Yeah, C jars, we're celebrating our 10th year this year, which is kind of astonishing to think that we've been doing this for a decade. C jars is an effort to really modernize our capacity to do research and statistical reporting. And what does that mean in practice? You know, we look at other types of areas that we do research on. Healthcare has been dramatically uh transformed by the availability of electronic health records, education research. Uh, you know, we have incredible breakthroughs being made in terms of thinking about the effectiveness of policies, teacher value added, a whole range of things that's been built off of the fact that we have electronic kind of administrative records for education, labor markets, and labor market research has been transformed by the availability of quarterly wage records that researchers are working with. And I wanted to do something similar for the justice system. You know, my interests really lie at the intersection between the socioeconomic reasons why individuals might engage in crime, as well as the socioeconomic ramifications of criminal justice. And when you look at the available data that's out there, a lot of it kind of puts crime and justice-involved individuals often kind of their own silo by themselves from an analytic standpoint. So you have things like the UCR, uh the Uniform Crime Reports, or NIBERS, the more recent reincarnation of it, or um the NCRP program, the National Corrections Reporting Program. A lot of these things track the state of the criminal justice caseload onto itself, you know, the number of people entering the system, the number of crimes committed, you know, the number of arrests being made. But you aren't really seeing the dynamic as people are moving through the system. And that's just because there's no central repository that keeps track of all of these different types of interactions that we have. So the idea about a decade ago that we had was to try to build something like this where we could say, okay, regardless of uh jurisdiction or time, like let's try to get all of the arrest records, all of the court prosecutions, all the periods of correctional supervision, whether institutional or community-based, and put that into like one combined data set so that you can track somebody as they moved through the system and see the repeated times that they're interacting, you know, whether within a jurisdiction or across jurisdictions. And going a step further, let's make that linkable with things like employment earn and earnings data, mortality records, household information. And we were pursuing this through a partnership with the US Census Bureau so that you could really kind of integrate these types of data and overcome these hurdles that we've had of just having kind of crime and justice as its own siloed topic by itself, rather than thinking about it in kind of the holistic sense of the challenges that people face across the whole life cycle.

SPEAKER_00 9:06

I you and I have done a number of different presentations for audiences together. And, you know, you present present uh CJRs and I present the real-time crime index. And I'm always amazed when you talk about the difficulty to achieve this, because it's the exact opposite of what we're doing in the real-time crime index. We want data quickly, we want it immediately, we want it as fast and close to real time as possible. You're doing the opposite. Can you talk through what a challenge it is to get some of this data?

SPEAKER_01 9:31

Yeah. Okay, so thanks for that. Anyway, so the procedures, though, like we aren't the first people to think of wanting to do this and having ideas of trying to pursue this kind of research agenda. The reason it doesn't exist is that crime and justice data in the United States lives in a very federated environment or system. And so what that means is practice is that often it's state and even local jurisdictions like counties or cities that are producing and maintaining their own uh case records of who's processing through their system. So that means like thousands of different types of data infrastructure systems, thousands of different types of legal environments that you have to navigate to figure out how to get access to data, and then a whole myriad of ways that information is being stored, whether it's like one long kind of like text file describing a case or delineated fields that explain specific aspects of a prosecution or something like that. And so, you know, what the bulk of the work is for doing CJRs is going out and building relationships with agencies across the country, navigating those administrative and legal hurdles so that we can ingest local jurisdictional data and then having a team of data scientists go through and work on processing that into a common format so that, you know, when we have one variable that says, okay, this was, you know, the sentencing date or this is the offense description, that means a common thing across all these different jurisdictions. And so there have been a variety of kind of tools that we've tried to develop to help make that processing more straightforward and easy. But, you know, it's it's pretty challenging. You even look at federal efforts and a lot of federal data collections that might not even be as I don't want to criticize federal agencies, but you know, you take something like the NCRP, the National Corrections Reporting Program. This is trying to do kind of something kind of similar to CJRs, collecting information on who's moving through prisons, uh state prison systems uh year by year. And there where with a smaller scope of just looking at prisons and just working with state agencies, not having to go to counties or uh municipalities, they don't have all 50 states participating consistently in the program. And uh I think that just set speaks to how hard it is in a system where there can't there isn't mandated data reporting to try to work within the constraints of agencies to get cooperative participation and and share this information for research purposes.

SPEAKER_00 12:12

Without throwing anybody under the bus or or saying talking about difficulties, has the environment gotten more difficult for acquiring data in the last few years, or is it something that it's sort of it's at the wonk level, so it doesn't really rise to a level where it might get more difficult?

SPEAKER_01 12:28

No, I definitely don't think this is at the wonk level. There are things that have made it easier and made it harder. So on one end of the spectrum, you have a number of states that are passing things like clean slate laws. And uh, whether by intention or by omission, often these end up precluding researchers from getting access to criminal records, which can makes it really difficult for us to do things like understand what the impact of something like an expungement or a record clearing permission might be. Because if we can't actually get the data in the first place, then we can't understand how it's actually whether or not it's accomplished the aims that it's set up to do of improving access to labor markets or something like that. I think on the other end, there are agencies, and this isn't a recent phenomena, this is something that's been known for decades, that feel burned by sharing data and access to their records to researchers or reporters. And given the political nature of running these agencies and the sensitivity around topics of crime and justice, there's real reticence to sharing that kind of information externally with partners. And so there are definitely lots of jurisdictions that we're working with very happily, and I think are it makes kind of the work that we do possible. But there are definitely a number out there who either because of their desire to protect the privacy of justice-involved individuals, or because of their concerns around self-preservation and not wanting kind of skeletons in the closet or data to be misconstrued to come out, that it it's just very hard to get to kind of assigned legal agreements so that we can work with the data.

SPEAKER_00 14:21

Is there a solution? I mean, it the states are doing too good of a job of cleaning the slate. Is there a sort of middle ground where you can protect the privacy and actually expunge a record while also enabling researchers to figure out whether or not these things are actually working?

SPEAKER_01 14:38

I mean, yeah, I think so. I mean, I think you can write laws such that uh records don't have to be purged by researchers. We worked to help develop a piece of model legislation that states can look into adopting that make it more explicitly clear to agencies that uh this kind of information can be shared with researchers and it would be used in a um secure and protected way, such that the desire to protect the privacy of citizens is still being honored. Uh, but yeah, I totally think that this is something that could be navigated. I just think that in the desire to affect change quickly, this is kind of one of the wonky things that people didn't think about the implications as they were passing laws. And it might be a good opportunity to make some uh fine-tuning.

SPEAKER_00 15:31

So let's talk about some of the cool tools you've built. I want to start with Joe. What is the Justice Outcomes Explorer and where can people find it?

SPEAKER_01 15:39

The Justice Outcomes Explorer is available at Joe, joe.cjars.org. Um, and it's a data dashboard that tries to curate information that we've learned from running the CJARS project. To take a step back, so we've kind of collected tons of information, probably a couple hundred million uh criminal justice events from over 40 million unique individuals in the United States. The data comes from over 40 states around the country. That's sensitive data. You think about that, and then adding in, you know, all of those people's prior and future employment outcomes, information that they report to the federal government through household surveys, their health insurance status, a whole range of like I think fairly sensitive things. The mechanism to get access to that is appropriately arduous and long. So you have to kind of apply to the federal government to get access. You have to work within a secure data center. The whole application process usually takes around 18 months. You have to go through a full background check. You can't ever have data on your local computer. You know, there are all these restrictions in place. And what that means in practice is that a lot of people don't end up working with C jars who might have questions that they want to answer. You know, I don't expect reporters or politicians or policy analysts to go through this whole process to try to get access to C jars. And so what we were trying to do with Joe is try to create something that could be brought out of this secure environment to inform public conversations around the justice system and the effectiveness of policy. And so what we did is take all this linked microdata, produced aggregate statistical series. So, you know, the average share of inmates leaving, let's say, a prison in Texas in 2010 who are earning above the poverty level. That's an example of statistic that's available on Joe. And then create kind of a user-friendly dashboard that people can interact with to kind of draw out kind of what information that they're most interested in. And so this creates new information that's in the public domain on a variety of dimensions. We're getting socioeconomic outcomes that we previously haven't had consistent reporting on or available, any available reporting on for a bunch of different justice and all populations, people in prison, people facing felony charges, uh, misdemeanor charges, parolees and probationers. We're also getting consistent definitions of uh recidivism. Recidivism is like such a going back to your question, like what causes people to engage in crime, you know, recidivism feels like such a basic concept. It's like, oh yeah, like we all know what recidivism is, but there are a bunch of different definitions of recidivism out there. And I feel like every agency picks the one that makes them look the best. And so if we're trying to look at like what policies are effective, we need to like be a little bit more precise on what this like recidivism concept is. Like, is it returning to prison? Is it facing a new felony charge? Is it any type of arrest? And here we're like putting all of that information out there so that you can pick your favorite definition and then look across all of these different agencies to see who's doing better than than other peer agencies or or who could potentially improve. And that can help inform like where we look to try to find evidence of effective policies. So, anyways, getting back to your question, Joe is something that we are really excited to put together to try to bring CJRs into the conversation for more people who wouldn't go through the the process of having to get access to the secure microdata. And um, we're really excited about it. It's publicly available online.

SPEAKER_00 19:34

Yeah. So is there a mechanism for sort of full state coverage? So some state like Texas, which is the same with crime data, has really good coverage. And then I look and I'm just looking at the map and I look at Louisiana, which I'm where I am, is uh not unexpectedly very poor coverage. So do you think there's a way to get to where there's sort of uniform coverage or better coverage in the place that don't have really good data, or are just the states that are good are the states that are good and it's just it's not gonna get better?

SPEAKER_01 20:04

Yeah, I mean as an organization, I feel like we've accomplished a lot with the resources that we've had. Um part of that is that we are like really strategic and intentional on where we engage. And so places that have, as an example, like a state administrative office of the courts where they've already put in the infrastructure to collect all of the county level court data into one statewide office that we can go to rather than having to negotiate with 50 to 100 county offices. Um, we prioritize those because we can get more data for fewer hours worked relative to like a place like Louisiana where everything is just completely decentralized. It's something that we hope as we grow as an organization, we have more capacity to take on. And as we build more experience, things like ingesting all of these different data formats becomes easier as we kind of there are only so many different reincarnations of like a court record. And so as we get more experience, we we kind of get a better sense of the lay of the land of not having to learn completely new ways of organizing information that we're getting. So I think I think there is like promise, but it does our job would be a lot easier on two with if there were kind of two changes. One is if states invested more in developing kind of consistent internal reporting mechanisms, especially statewide offices, for uh local agencies to report up to. And the second would be just like sufficient investment in general in terms of internal capacity. And so often, you know, we're we're getting pushback from agencies when we go to ask for data just because like they literally don't have someone there who can produce a data extract for us, given the zillion other things that they have on their to do list that they're behind on. That's just a really common issue that we run into. And it's not that. Anyone doesn't want to see our project or other types of research not to succeed, but it's just like you know, we have a part-time person who's here and they've got a bunch of other IT issues that they have to pay attention to and not just this random academic research request.

SPEAKER_00 22:17

If you could wave a wand and magically create a data set or a partnership or something, is there like a, you know, Moby Dick that you've been chasing that you really want that you just can't get, or you sort of happy with the way things have been progressing?

SPEAKER_01 22:32

One of the things that I feel like we haven't done enough of and that we should be investing in more is engaging with industry associations. So like the National Sheriff's Association or something like that, and going and and really investing and building the relationships with these groups to establish trust and credibility in what we're doing. I feel like a lot of these agencies are probably bombarded with a lot of requests to get data out of them from researchers. Um, I think that's facilitated by um, you know, electronic requests. I'm sure with generative AI, it's gonna get even worse because it's just gonna be really easy to just buy off, you know, a thousand emails to a bunch of different places. And so I think in response, a lot of agencies who don't have enough capacity just like shut down on the entire process. And so I think potentially working through industry associations, we would be able to kind of help better signal that like we are we aim to be like a really professional organization, that we made serious investments in terms of the safeguards, both from a technical and a policy standpoint, that they should trust their data with us. So I think I think that's like the realistic thing. That's not waving a wand, that's just like we need to hire X more bodies. And so as a grant funded organization, that just comes down to finding the right funding and a partnership that we can pursue to do that.

SPEAKER_00 24:05

Moving on to the next tool that I wanted to talk about is the text-based offense classification tool. Because this is really cool and and um it it provides for the ability to essentially figure out like standardize what offenses are and using just tech. I mean, what walk me through this tool.

SPEAKER_01 24:23

Yeah, yeah. So this is something we developed a few years ago, and it was like completely out of need because um, you know, we were facing this influx of data, and you know, the the records we get from agencies, they aren't designed for research purposes. These are case management tools. And in the context of crime, you often want to have like maximal flexibility. I think if you asked uh a police officer or a prosecutor to go through and you know, check one of 68 crime types, like what occurred here, and not like put a narrative description of what happened, you would get a lot of pushback because every kind of situation is kind of different and it's not so simple as a categorization. But from a research standpoint, that's what we have to do. You know, we can't individually go through 250 million records and decide how we're gonna say, oh, this is a robbery or not. So, in partnership with Measures for Justice, we uh developed a machine learning tool. Measures for Justice had invested an incredible amount of time and energy to hand coding something like 300,000 offense descriptions. In our database, we have like four million plus unique offense descriptions. It's just because there's a lot of contextual information, a lot of abbreviation. Sometimes the laws are being cited, sometimes there's just misspellings. I think there are like a thousand different ways of spelling marijuana in our database. And so uh they had spent the time and energy of having developing a schema and having kind of trained professionals go through and like map the that, each kind of a description in their data to a specific category. And so then we built on top of that a probabilistic prediction model using machine learning methods to say, okay, if you have this kind of collection of characters or um words in a description, then that probably means that this is a drug possession charge, or this is a robbery charge, or this is, you know, a larceny charge or something like that. And then applied it at scale to our data set. And because we had built this and we knew that there were a lot of people out there working with their own administrative records, we figured, you know, why not just put this tool online, you know, not our underlying microdata, but you know, anyone could process their justice data through talk so that we're moving towards having a more consistent uh approach to how to manage the front end of data wrangling of administrative records before it gets to the research phase. And that's been really successful. I think there are like a couple thousand times, a couple thousand different data sets that have been run through talk now. And I think it's probably one of the things I can point to that has the broadest public engagement. And it's really interesting. We see all the people who register and in some of these federal agencies, we see people register for it. Um uh state agencies, uh researchers, you know, it draws in a wide range of people who are kind of interested in seeing how this works.

SPEAKER_00 27:23

That's great. Is there a tool that's sort of the next tool or one that you want to build but haven't quite gotten around to yet?

SPEAKER_01 27:31

One of the ones that we have done, uh but we haven't put into the public domain yet is we have this entity resolution algorithm. Um, that's a fancy way of saying it's like a probabilistic fuzzy matching tool to figure out who the same people are. So let's say that there's you know a crime record in Washtena County where I live, for like Mike Mueller-Smith. And then somewhere else in the state of Michigan, there's a Mike Smith. And then in Chicago, there's a Michael Mueller. It's like, are these the same people or not? And using kind of the collection of personally identifying information that we have trained another type of machine learning model on for some jurisdictions where we have an ID that's verified by fingerprints. We've figured out a way to do, I think, a very good job of identifying who we think are the common, which records belong to the same person and when which records where we have to be like, we think these are two different people. I think there are a lot of use cases of having something like that online. There are a lot of other approaches that are that are out there that use more, it's called deterministic matching or kind of set rules. And I think that, yeah, I think that this could be really useful, but it's a really large tool and we don't have like a server set up or anything like that to do it.

SPEAKER_00 28:52

How does AI factor into all of your work? It feels like a lot of the recent developments could just expand it tremendously. Is that something that you guys are working on?

SPEAKER_01 29:00

Yeah. So we have had some data partners who have confirmed that we use no AI in our servers because they're not comfortable with that. And I think that's right. Um, I think there's a real tension on when you're working with sensitive and secure data. You don't want to unintentionally leak private information to public domain. So, like our production servers are essentially severed from the internet so that there's no unintentional data breaches occurring. That's kind of effectively the same kind of environment that we do research in as well. Um, so in those cases, you you can't really use generative AI to support consuming the data. One thing that we are kind of processing the data, a few areas that we've talked about doing and integrating AI is, you know, we probably don't have, given the number of researchers using C jars, we probably don't have the right level of like customer support, so to speak, for what we're doing. And so I could imagine like feeding in all of our documentation about C jars. You know, we have a data documentation PDF that's like several hundred pages p long. And I can tell you, I think like maybe one person has read all the way through. And I think it's just on our like yeah, no, it's Jordan. It's like nobody actually reads this thing, but like I think if we and as a result, we get a lot of emails from researchers for things that are just already available online. Um, so if we potentially created some sort of like interactive AI agent that could help answer questions and direct people to resources more quickly, I think that could be really helpful. We do a lot of web scraping for jurisdictions that publish records online. And in those cases where data is just kind of already in the public domain, I think AI could potentially help us harvest that information faster and parse it faster so that it's more easily ingestable. Uh, but I think like for a lot of the data, and once it crosses into the domain of being on the secure servers, options are kind of limited because the privacy and the security of the data comes first.

SPEAKER_00 31:12

So my last big question is having done all of this work, putting all of this together, do you have any sort of big overarching lessons that you've learned? What have you learned that could impact policymaking and make our approach to crime and the criminal justice system smarter through all of this?

SPEAKER_01 31:29

Yeah, I mean, I think that there are a lot of options available as we think about combating crime. And there are a lot of, as we think about like the array of tools that we can deploy to combat crime, we should be thinking kind of in a broader sense than just like do we need to up the number of police that we have or have more or less incarceration. Um, I think we're um thinking potentially about like a broader set of interventions, including things like health coverage, income support programs, like what the ramifications of those types of things are. I also think that we've all grown to appreciate that there are wide-ranging consequences for justice involvement and that we have to think about the nuances of the way that we operate the system, that it's not just like the effect of incarceration or not. You know, there are millions and millions of people out there living with criminal records that have never been to a prison yet, face severe barriers on kind of reintegrating and reestablishing their lives to lead productive lives that don't turn to crime again in the future. And so I think, at least in my mind, it's really hard to separate these domains that, you know, people live complex lives and we should approach the way that we combat crime and the way we run the justice system accordingly. And Mike, what what is next for you? Oh, I I don't know. I mean, I think like C jars is is just it's it's the gift that keeps on giving. You know, we're in the process of trying to compete for a variety of different federal grants that will kind of apply the the learnings that we've had from CJRs to other types of federal data collections that we're really excited about, you know, continuing to grow C jars and kind of see it on a path towards being more financially sustainable, you know, and just continuing to get back into the world of research. You know, I started this from a viewpoint of like there were questions I wanted to answer. And, you know, obviously I was on a tanger track, and so I was still pursuing research, but there is a big kind of tangent I went down in terms of, you know, getting seniors off the ground and and and set up and sufficiently uh staffed uh with leadership to be a little bit more independent. I think we're starting to get to that phase, and I'm excited about you know diving back more fully into research. And so that's that's I think what's ahead of me for now.

SPEAKER_00 34:04

All right. Well, good luck. The website is cjarz.org, correct? Yep. Yep. So check it out if you're interested. And Mike, thank you so much for joining the show. I this is it's such great work and it's such a unique project for making our criminal justice data work together. So thank you for all you do. Thanks, Jeff. Thanks for listening to the Jeffalytics Podcast. Be sure to subscribe and to learn more, head on over to ahdatalytics.com for more information and previous episodes. If you like what you heard, please leave a glowing review, which will help others to discover the show. Until next time, I'm Jeff Asher.