Homeless Issues Committee - Regular Meeting

So I'm going to formally call to order the meeting of the Community Justice Council Pretrial Subcommittee, of which I'm the chair, Judge Nicholas McNamara. It was possibly going to be joined with the broader Community Justice Council, and they're present as well. But we'll take a roll call, I guess, for the Pretrial Subcommittee first. We have a quorum for that.

A guard here. Clark Bernhardt here. Cron.

Here.

McNamara here. Middleton here. Morgan here. Okazaki here. Olison here. Ozan. Pierce here. Splinter is excused. That is quorum.

Okay, good. So we have quorum so we can Have a meeting. I'll try the role just for CJC.

Agard? Here. Barrett? Sheriff Barrett? I can't see folks online, so I want to make sure I'm giving people a chance. Dillard was excused, but I thought I saw Jerome. Someone else is using his. Okay. Excuse Dillard. Highland, excuse Laffenberg. Madison. Like there is excuse Moyer. Yep, sorry about being a little late. Middleton here. Miles, I'll say Miller here. Ozan is planning on coming. I've received indication, but I don't see him. Pedersen excuse push excuse like lightener here. We do not currently have quorum. So CJCPT has quorum, so you'll run the meeting.

So I'll welcome everyone else and the community generally. Since we do have quorum and it's an official meeting, we would normally start with approval of minutes from our last meeting from June 4th. BUT BECAUSE OF ITEMS ON THE CALENDAR AND THE JOINT SESSION, I WOULD ASK THAT WE DEFER APPROVING THOSE MINUTES TO OUR NEXT MEETING UNLESS THERE'S AN OBJECTION. ANY OBJECTION FROM OUR PRE-TRIAL SUBCOMMITTEE TO DEFERRING THE APPROVAL OF MINUTES? NO OBJECTION. OKAY. WE'LL DEFER THOSE THEN. AND WE WILL BE STARTING WITH PRESENTATION. I UNDERSTAND THERE MIGHT BE A PERSON REGISTERED TO SPEAK REGARDING THE PRESENTATION. Clark Rogers, was Clark present? He's not here right now. I don't know if he's online. Okay. So maybe he was thinking to come later. I don't know. Well, then, Colleen, I'll pass it to you to introduce our speakers. Sure.

Great. Thank you. And please come up, Professor Greiner, from the Access to Justice Lab, and Ryan Halen, who is a data analyst with Access to Justice Lab from Harvard Law School today represents a decade-long partnership with the Dane County CJC and the Access to Justice Lab in various forms from early conversations to early MOUs, to early data calling to our initial report. And now Dr. Greiner and Ryan will speak about the finalization of the other sites around the country that we're also looking at the public safety assessment as well as cash bail and the system of relief. So without any further ado, our visitors, I welcome you to Dane County. It's a hard day for a lot of folks here. So I just hope everyone gives grace to that and we can still have a conversation.

Thank you very much for the invitation to be here. It's lovely to see so many familiar faces and to meet a bunch of new folks. Colleen, as you just mentioned, this is a long-running partnership among the CJC, Dane County, Arnold Ventures, who funded our work, as well as some initial seed funding for the county and the Access to Justice Lab, and we're delighted to be here. So thank you so much. As you mentioned, Colleen, the first meetings for this project took place over 11 years ago, so it has been a long-running effort. The setting in which this study took place that we're going to present the results of and further research about... is predisposition in custody, first appearance hearings, and where key decisions are made that affect the course of a criminal prosecution. And the key decision that we'll be studying today is the decision to assign a signature bond, which means that an arrested individual can, for the purposes of this particular prosecution, go free without posting money with the court or cash bail decision. in which the individual does have to post money with the court before being able to walk free for the purposes of this particular case. And there are consequences, quite serious consequences for both the individual involved and for the county in that decision. So there's risk either way in any decision that one makes. If one makes a decision not to impose cash bail, there is a risk of a potential figure to appear, new criminal activity, particularly new violent criminal activity. If there is a decision to impose cash bail, there will be an imposition of financial consequences or well, there will be an imposition of financial consequences either for the individual who has to post the cash bail if they're able to, or if not for the county who has to incarcerate. as well as socioeconomic consequences for the individual and the community as a whole. So it is a high consequence decision. And there has been a tremendous amount of interest in it, especially over the past 25 years in the scholarly community. One of the things to emerge from that interest in the scholarly community is the idea of having a quantitative instrument to attempt to provide information to the decision maker in this county called a commissioner that would potentially advise or guide the decision of the commissioner. I want to emphasize that so far, at least few, if anyone in the scholarly community or anyone serious in the implementation on the ground has proposed the idea of having the risk assessment instrument determine the decision. This is simply advisory information that the decision maker, the commissioner, is free to disregard if they so choose. And typically the instruments, the quantitative instruments are designed with the idea of providing information that can be used within the statutory constraints that the commissioner has to operate within. So for example, in Wisconsin, at the first stage, my understanding, and again, we have some former commissioner for 30 years here. Good to see you again, Commissioner Maurer. It's always good to see you. And Your Honor, obviously, you would know this as well, Your Honor. My understanding of Wisconsin law is that the first inquiry in a decision about whether to impose cash bail is failure to appear. And that's the critical factor that is to be decided first. And then later in the process, if at all, concerns of public safety can be incorporated. And so the risk assessment instrument or the assessment instrument, whatever it is, must be flexible enough to be used within various statutory contexts. And this one, the public safety assessment was designed to do that along with some other things. And for those of you that are unfamiliar with the public safety assessment, it consists of two one through six scales that purport to classify individuals according to the risk if released. of committing, in the first scale, failure to appear, in the second scale, being rearrested for a new charge that could involve jail, and a 01 flag, either the flag is there or it's not there, that is supposed to indicate an elevated risk of new violent criminal activity if released. And so it's a one through six scale for failure to appear, one through six scale for new criminal activity, a zero one flag for new violent criminal activity. These are the inputs from this risk assessment instrument, or excuse me, the risk assessment instrument uses. All but one of them consists of criminal record information, if any, so past criminal history information, the one that does not as a demographic factor in its age at the time of the arrest. So there's no race, for example, in this risk, there's no race code in this risk assessment instrument. The way one uses the PSA is to marry it with a jurisdiction-specific decision-making framework that grids the result of the one through six scale for FTA and the one through six scale for new criminal activity, the NCA scale. into a set of potential recommendations or classifications of decisions. So for example, if both the risk assessment for NCA and the risk assessment for FTA are low, the individual will be classified in the top left box of this grid and the color green, and the recommendation would be a signature bond with no conditions. And then the green gradually turns from light green to yellow, to orange, to red, As the risk assessment scores increase, the 01 NVCA, New Violent Criminal Activity, flag operates outside of this grid, but it is part of the process by which the PSA scores are transformed into a recommendation for the commissioner. And so that's how the overall process works. And the idea is that assessors or a computer program in some jurisdictions will calculate the risk assessment instrument score and provide it to the commissioner and the commissioner can do, and in the case of Dane County to the prosecution, good to see you again, and the public defender's office, and I think we have several, at least one member I know we met with you, Peter Middleton, earlier this morning, to the parties and the commissioner to inform the colloquy that they have and the decision that the commissioner makes. So that's how the risk assessment instrument works. Here is a redacted example of one of a PSA output. There are the outputs of the PSA scores. You can see that this particular individual, for example, did not have a 01 flag. So there was no elevated risk of NVCA. There was a four on the new criminal activity scale and a three on the FTA scale. And so there's that information blown up. Here are the inputs and so that the commissioner and the parties have access to the inputs to the PSA as well. And then there's the recommendation down at the bottom that came from that grid, that colored grid that I showed earlier. So what are the hopes, or what were the hopes for the PSA-DMF system? They were to reduce some or all of the three criminal outcomes that I mentioned earlier, FTA, NCA, and VCA, as well as to reduce racial disparities that are operant everywhere in the country in the criminal justice system, and to try to reduce the number of predisposition incarceration days. In other words, to try to reduce reliance on incarceration. And so the county made the decision to introduce the PSA and begin its use, and that use began in 2017, but at the same time, it created the partnership with the Access to Justice Lab to try to figure out whether that introduction worked. Did it actually produce any of these results that I just put up on the slide? Would it actually do so? And the question then became, well, how would we know whether the introduction of the PSA had, in fact, produced or not produced those results? And we borrowed an idea from medical studies, from the approval of new drugs and medical devices that the FDA requires the drug companies use. And basically the idea is that we conducted a randomized study. And so we randomized whether the commissioner and the parties would have access to the PSA printout in a case or whether they wouldn't have access to it. And the randomization assured that all of the case, the set of cases for which the PSA was present would be identical to the set of cases in which the PSA was not present. except for the presence of the PSA. So any differences that came between those two groups could be safely attributed to the presence of the PSA, because there were no other relevant differences in the sets. And this is an old idea, again, one used, required in drugs and medical devices, and here is the basic schematic that we implemented. We implemented this from 2017, the middle of 2017, until the end of 2019. The county was then kind enough to undergo a process of producing data from various offices, especially the clerk's office and the prosecutor's office produced data that we analyzed and spent a long time analyzing. In January of 2024, Ryan Halen and I appeared before this group remotely and reported results that I'll repeat today, but in short, that the PSA did not produce any of the benefits, nor did it produce any of the harms. that either its proponents or its detractors had suggested that it would. And so neither the hopes nor the fears for the PSA ended up being well-founded. And at that time, we had hypotheses as to why that was true, but we could not draw conclusions as firm as we would like. And the reason that we could not draw those conclusions, or at least that we advised waiting on those conclusions, was because we had two counterpart studies that were occurring in other counties. And we wanted to get the results of those other county studies before we could see whether they were the same or different as those in Dane. And that's why we're back here today, is to, again, refresh your recollection as to exactly what the results showed, our hypotheses as to why, and then what the other county results showed so that we could see Dane in the context of the broader picture of the studies around the country that we had conducted on the PSA. So again, these are the outcomes that we studied. They basically match the outcomes, the hopes and aspirations of the PSA. And these are the things that changed between this presentation. It's the conclusion of those other studies and the level of firmness with which we can represent whether we think we found the reasons why the PSA doesn't work. The additional studies that I mentioned took place in Polk County, Iowa and in four counties in Utah. These are all roughly outside of Salt Lake and they agreed to participate and we consider those counties as a joint unit because none of them are large enough on their own to produce a volume of information for us. And a key moment was with the involvement of an additional scientist, an old friend of mine from graduate school. His name is Kosuke Amai. And Professor Amai is a world-renowned statistical methodologist who has won multiple awards as an applied statistician and a political scientist. And he agreed to come work on these problems with us. The problems that I'm identifying are that at the time when we began this effort in 2017, The math wasn't there. The statistical framework, the principles, the statistical framework was not there in order to make the best use of the data and to study the questions that we wanted to study. So in partnership with Professor Amai, we created the math. And we have now published six papers or posted six papers in statistical journals. This is perhaps the key paper. Those were two that are already in publication. This is the one that has recently been posted that allowed us to make the conclusions that we're making today. And so again, what are we going to do? We're going to compare results for when the PSA DMF system is present among the parties and for the commissioner and to the cases in which it is not. So comparing these two sets of cases. What happened with respect to commissioner decisions? This, by the way, we did report two years ago. So for those of you who were here, you will know that there was a moderate effect on commissioner decisions, but no effect on anything else. So what did that look like? were for commissioner decisions. Here is a chart, both of these charts appear in the, two charts, both these charts appear in our report. And what it shows, for example, is that when the NCA score was five, so NCA again is new criminal activity, so this is a relatively high NCA score. the Commissioner's decisions moved, and they moved in the direction that one would expect. In other words, when the PSA was present, there was a 30% use of imposition of cash bail, When the PSA was not present, there was an 11% imposition of cash bail, and this is for the set of cases that scored NCA 5. So that's to be expected, that it moved in that direction, and that difference is statistically significant. In other instances, there was no difference, and this probably shouldn't surprise us. When the FTA score, so failure to appear score, was extremely low, There was only a three percentage point difference in the use of signature bonds. So there was a three percentage point increase in the use of signature bonds when the FTA score was low. That is in the expected direction. It's not statistically significant, probably because there's very little room for it to move. When you're already providing 90% of people in a certain class signature bonds, there's not a lot of room left. You basically can't get higher than 100%. And so there wasn't a lot of room there to move. And I will note that nationally, as compared to national criminal justice systems, this county uses both citation versus arrest and signature bonds at much, much higher rates than we have seen anywhere else in the country, quite frankly. So You can see that, for example, in a statistic that's not reflected in the slides, around 70% of individuals that reach in custody, first appearance decisions are charged with felonies. Only 30% is the highest charge misdemeanor. In most jurisdictions, that's reversed. In most jurisdictions, 70% of people reach in custody first appearance hearings are highest charges of misdemeanor. 30% are felonies, and that's because you all use so much cite and release here. And in addition, your signature bond, other jurisdictions call it release on recognizance, ROR rates are quite frankly astronomically high when compared to those of other jurisdictions around the country. So there wasn't a lot of room to move when the risk assessment instrument scores were low. Overall, what happened was that, to our surprise, the presence of the PSA increased the use of cash bail by a very small amount. And it was statistically significant, so it's there, but it's about 2.5 percentage points, and it's certainly not huge. But we didn't think that was particularly what we had expected to see. And it wasn't what the folks that came up with the PSA had expected to see. They thought that there would be a reduction in the use of or the imposition of cash bail obligations when the PSA was present. And that's what they hope for. It's not what it's turned out to be either in this county or anywhere else. Quite frankly, there's just been very little. There have been very little overall effect. So again, this is statistically significant and real. It's not huge. And so it's up to the county to decide whether it's policy relevant, I guess is my point. With respect to FTA, NCA, and VCA, as we reported previously, there is no effect overall. We have new and more accurate measurements of these quantities that are different from the ones that we presented before, but the conclusion is the same, that there's no difference. The reason why is because of the creation of the new statistical techniques told us that what we needed to do was separate people who were on their first arrest within the study period from people who were on their second arrest within the study period in other words within the study period they were arrested once released and then re-arrested and then third arrest within the study period again all within that period from 2017 to 2019 again this is not having arrests before 2017 where we treat it as a background variable and as an input to the psa These are first, second, and third arrests within the study period. The math that we created told us that we needed to treat those cases the right way. The result was the same. There was no differences. When we did the right thing, we got the same conclusion as when we did the wrong thing. which is that there are no differences when the PSA is present and when the PSA is not present. Here is a comparison of the green or yellow and the red. The green is the first time within the study period that a particular individual is arrested. The yellow is the second time within the study period the person is arrested. The red is the third. There were too few of the fourth, the fifth, and so on, so they don't appear in the graphs. There's no differences for them either. And this is the result for NCA. Again, there's no, these differences are tiny. The only reason you see them as all is because we blew the scale of the graph up. If we added on a zero one scale, you wouldn't even see these things. And again, no differences. And overall, for all three of the jurisdictions we studied, Dane, Polk, and the four counties in Utah, there was no pattern of statistically significant differences. There were two actually that were statistically significant, but one of them was in the direction that nobody wanted it to go in. whereas the other one was in the direction that people wanted it to go in, which means things in one case got worse with the PSA present, in another they got better, but they got worse in a different dimension. So one of them got worse in the FDA and they got better in MDCA and they were two different jurisdictions. There's no pattern. This is just what we call statistical noise. There's nothing going on here is the overall. So overall then the results for Polk and Utah were the same as those in Dane, which means that there was no effect of the presence of the PSA in any of the jurisdictions. There was no policy relevant or really statistically significant effect. This is just the kind of thing that you would expect if you rolled a lot of dice an awful lot of times. you would get this kind of stuff. You would get this kind of stuff. So there is nothing, the news here is that there is no news on the FDA, NCA and BCA. And it's true in Dane, as we reported two years ago, it's also true in Polk, turned out to be true in Polk and Utah. What about predisposition, number of predisposition incarceration days, which is obviously an issue of tremendous importance in the community. And the answer is no effect. And so the way to read this graph is that blue is the set of cases in which the PSA BMF system is not present in the Commissioner's hearing room. Pink is when it is present and purple is overlap and everything's purple. So if there had been a difference, we would have seen some visible blue and some visible pink. Now, that's obviously not a statistical test, but we ran the statistical test, and it told us the same thing that this graph visually tells us. It's all purple. There's nothing there. There's no difference in the number of predisposition incarceration days when the PSA DMF is present as opposed to when it's not. Measures of racial fairness, instead of light blue and light purple, we have red and blue. Red are tests of racial fairness. Blue are tests of gender fairness, which is something that in the criminal justice system, most people quite frankly don't care about. But we studied it anyway. And things to the left of the vertical line at 0.05 are statistically significant. And we ran an awful lot of tests. We ran an awful lot of tests. And so the point of this graph is not to say is anything to the left of the statistically significant line, it's is there a pattern of anything being to the left of the statistically significant line because again we ran so many tests that some of them are accidentally going to show up as statistically significant that's just you know sometimes if you roll if you roll a die enough times you will get an unusual looking run of say twos in a row just because you rolled it off a lot of times. And so that's the equivalent of what we did here. And there's no pattern. And what pattern there might be there honestly looks like it's more with respect to gender than to race, but there's no pattern overall. And so this is what led us to conclude that whatever racial disparities are in the criminal justice system, and they are, they're not made better or worse with the presence of the PSA. They're just unchanged. Okay, and this is all what we, essentially what we reported were more accurate and were more scientifically rigorous, but essentially what we reported two years ago. No effect on any of these key decisions. So what's the difference? Neither the hopes nor the fears are realized. What's the difference? The difference is in why. And two years ago, we had three hypotheses that we were thinking about. One of them is that the changes in the commissioner's decisions were not large enough. In other words, the commissioners did not follow the PSA recommendations with sufficient fidelity, they didn't follow them enough for the PSA to make any difference. We were skeptical of that decision, that explanation at the time two years ago, and now we are extremely skeptical of it. We don't think that was the answer. That was the reason. The second has much bigger implications for the criminal justice system as a whole, which is the suggestion that bail neither prevents nor deters misbehavior within a predisposition misbehavior. It neither prevents nor deters failure to appear new criminal activity. In other words, the imposition of cash bail neither prevents nor deters individual, prevents via incarceration nor deters via a fear of monetary consequences, failure to appear in new criminal activity or new violent criminal activity. That was our second hypothesis. And our third is that the PSA classifications were not strong enough to do any good, that they were not separating individuals on risk well enough, okay? So those are our three hypotheses and this is where the changes come. So the reason why we are fairly certain that these culprit was not changed in commissioner's decisions was because in Polk, commissioner's decisions changed an enormous amount and we still saw no effect. there was an absolutely startling amount of change in the direction that the PSA suggested in Polk County, and they still saw no effect there. So for example, when the NCA was one, There was an over 30 percentage point movement, 35, it was around 35 percentage point movement in the use of OR release, their word for signature bond. Now, if I recall correctly, Ryan, those rates, initially those rates of OR release were fairly low. Is that right in Polk?

Yes, unlike Dane, where we saw the majority getting signature bond release in Polk and also in Utah, the overwhelming majority of cases saw some sort of imposition of bond or bail in order to secure release.

And so there was a huge movement as a result of the imposition of the introduction of the PSA. But again, here are the results in Polk and there's nothing going on, especially with respect to NVCA, New Violent Criminal Activity, which is the major public concern. What may be going on there is not in the direction we want it to go anyway in Polk. In other words, there's an increase in FTA as opposed to a decrease in FTA. So we do not attribute these results to the quote-unquote failure, and it's not clear it would be a failure anyway. It could just be the commissioner's decisions, right? That's what, again, it's their discretion to follow the PSA DMF recommendations with quote-unquote enough fidelity. What about bail not preventing FDA, NCA, or NVCA? Here I need to introduce some terminology, and it's statistical terminology. It's not a comment on any one individuals, okay? The terminology I want to ask you to follow here is a set of categories of an individual. Imagine there is an individual appearing in an in-custody, in-person first appearance hearing. And they're in a simple as possible way of representing this individual could fall into four categories. One category is safe. And what we mean by and again, these are statistical labels. What we mean by that is the individual will not FTA if given cash bail and will not FTA if given a signature bond. In other words, it doesn't matter what decision you make, they will not FTA. They will not NCA, they will not NVCA. They will commit none of these forms of misbehavior, okay? We're gonna call the individual safe. Obviously, you can't know this with certainty in advance. Okay, so this is a statistical construct, right? Okay? What about a preventable individual? A preventable individual will commit some form of misbehavior if, and do I have that backwards? That'd be bad. Yeah, I do have that backwards. Let me change that. I will have to change it later. The individual should have gotten this wrong. I've got this dead wrong and I'll give you a new set of slides. The individual will commit a form of misbehavior if given a signature bond Can't believe we didn't catch it. If given a signature bond, but not misbehave if given cash. In other words, cash prevents, the decision to give cash prevents the FTA or the NCA or the NVCA. Okay? That's a preventable individual. All right? Then a hopeless person, and again, this is a statistical construct. It's not a comment on any one individual. A hopeless person will commit the misbehavior regardless of the decision. They will FTA if given cash, they will FTA if given signature. There's nothing you can do with the bail decision to stop them. And then a backlash person who, for reasons we can explain, will assume do not exist, they will do the opposite of what you'd predict. They will commit FTA if given cash, but will not commit FTA if given a signature. So the deterrence decision backfires on them. Consider someone, for example, who in order to, is given cash, excuse me, if given signature would show up to all their court hearings, but if given cash goes out and borrows money from a relative and then gets a third job in order to earn the money and because of the third job doesn't show up to the court hearing or something like that, okay? We're gonna assume those individuals for the present banalysis don't exist. I can tell you what happens if we don't make that assumption, okay? So we're gonna have safe, preventable, and hopeless individuals, all right? According to whether the imposition of cash bail changes their behavior. The ones that you should focus on, again, reverse the yes and the no here on this chart, the ones that you should focus on for cash bail purposes are the preventable individuals. Those are the only ones whose behavior the cash bail decision can affect. For everyone else, it's pointless. It's pointless either because it doesn't matter, they won't misbehave, they'll do what the system desires them to do, or they will misbehave and you can't stop it with the imposition of cash bail. So the question becomes what fraction, and the ability to answer this question comes from the new statistical techniques that I mentioned earlier that we've developed over the past six to seven years. What fraction of individuals in the county are preventable? What fraction of arrested individuals who make it to in-custody first appearance hearings are preventable? And for NVCA, which is the one that we've highlighted here because the violent rearrest is the one that, again, tends to have the most resonance, the answer is fewer than 4%. Basically, there is a vanishingly small fraction of individuals that the imposition of cash bail can either deter or prevent misbehavior. And we're happy to take questions about that. That 4% turned out to be by far the highest fraction that we saw in any of our three sites for the new violent criminal activity, which is the right-hand column of graphs here. In the other three sites, that fraction turned out to be so close to zero that we couldn't distinguish it to zero for NVCA. And for failure to appear and new criminal activity, the fractions were higher in Dane. We think, again, that's because of the fact that y'all are 70% felony in custody first appearance hearings, as opposed to the other counties that are 70% misdemeanor, and so the fractions are much lower in the misdemeanor population. But even so, in general, around over 90% of individuals, cash bail has no effect on what happens. And so overall, especially if you focus on the violent rearrest, we think that a major reason why the PSA doesn't work is because it was designed to advise something that doesn't work. In other words, the PSA is designed to advise the imposition of bail, but if bail has no effect, then the PSA can't either. Okay, and so again, that's what this is showing here. What about the idea that the PSA classifications are not strong enough? Well, we think that's also true. And we think that we've found better measures of that over the past five to six years. Here's the graph that everyone currently uses. Everyone in this business, this sort of nerd out business that we represent and are very much of a part of, this is what they all do. They basically say, hey, for people who had an FTA score of one, what fraction of them misbehaved in this particular way? How about two? How about three? How about four? And does that graph go up as the number, as the risk score goes up? Does it increase? We think it should increase a lot. And Ryan, you were innovating a lot on this over the past five to six years. And for example, you thought, you came up with the idea, well, at some point, for example, it should go above 50-50, right?

Yeah, specifically that, you know, at the higher ends of the scores, one would expect that you should have a better chance of detecting someone who's going to fail than of detecting someone who isn't. But that was only ever true in exactly one outcome in exactly one of the test sites.

Which is about what I'm going to, which is the next graph. However, we'll get to that in a second. In other words, we think the score should go up like this. It should increase quickly. And they don't. And again, I have a slide somewhere. Wait a minute, let me get the right one. No, don't have that slide. Anyway, they all look like this. All the ones in Utah and in Polk and in Dane, they all look like this, where they go up, if at all, only slightly. And some of them look like shot ducks, where the graph looks like it's going to go up and then kind of gives up about halfway or three quarters of the way and then goes back down. So it looks like the duck's gonna take off a little bit and then it gets shot right in mid-flight and then it just gives up and dies. And so that's an indicator that things haven't gone well in terms of the strength of the PSA classification. However, we think there are better indicators. Remember this graph here where we define these categories, although, again, preventable has yes and no reversed. Safe people, you can do whatever you want, they're not gonna misbehave. Hopeless people, you can do whatever you want, they are gonna misbehave. Preventable people are the ones that you can prevent with the imposition of cash bail. We think that an assessment instrument should be higher when there is a higher fraction of preventable people. Because if the idea is high score means commissioner, signal, impose more cash bail, The reason should be because there's more chance of getting catching preventable people there. Does this make sense? You know, more preventable people, more cash bail is the idea. That's what a score should tell you. And the PSA does not. Here are the results of the classification of the set of preventable people across the scores. So the bottom 1, 2, 3, 4, 5, 6 goes up. as uh and then and the scores again you can see for example in the time the dane fta the top left graph you get that shot duck picture that i talked about where it looks like the duck is going to take off and then it just gives up right your honor is that because cash bill was imposed on the people that the score was sick not for this graph okay for this graph the answer is yes That's the one that just says, who got out and did they misbehave and what were their scores? This is the one that uses the more recently developed statistical techniques to find out, to estimate the fraction of preventable people. And that takes into account the decision. That's the advance that we made over the past five years. And that, it took us a while. So that's precisely in order to answer that question that we spent the past four to five years on whiteboards, is to be able to answer that question. I'm highlighting the results in Utah because we are extremely sure about the results in Utah. In Utah, we saw all of the information that the decision maker called a magistrate saw. And that's a necessary assumption that we have enough or all of the information that the decision makers saw to our statistical techniques for them to work. We are in conversations with Clark Ozaki, with Defender Middleton to try to make the same true in Dane. And we don't have all of it yet. And so we're hoping to get all of it, this information that the commissioner had at the time of the decision. Nevertheless, all of the results look the same. And so we are reasonably confident in Dane that these results are all true. We really aren't sure in Utah. And so what are the implications? Well, the implications immediately, bail doesn't prevent or deter. PSA classifications are not strong enough. Those two appear to be true. Now what? Implication number one, perhaps consider redeployment of resources if any are continuing to be expended away from the continued production of the PSA DMF system. It doesn't appear to be doing anything. Number two, perhaps consider reduced reliance on cash bail. That doesn't appear to be doing much either. And by appear is a euphemism. It's not. It's not doing anything. On overall, there are people for whom it does do something and it may be a very, for those people there needs to be cash bail. They're a very, very small fraction. Okay. Number three, we suggest that we should do the same sort of an investigation that we did here into levels of supervision. Because that appears to be the other major lever that the commissioner pulls at the time of a first in custody, first appearance decision. And we should investigate that fairly as well. We are grateful for the opportunity to work with you on this over the past more than a decade and are happy to answer questions.

Just on that last point, the level of supervision means like pretrial services. Correct. Like monitoring, are we doing breath testing?

Correct. Yes, Your Honor.

Questions from folks here? Todd, go ahead. What's the MDRC report?

There was a recent MDRC study, a relatively recent meaning something like the last four years, so recent in the scientific sense. And we only saw the slides for that. We haven't seen the actual report. I'm not sure there was one, but there was a presentation. Actually, Colleen may have it for you. that investigated the level of supervision, the effective level of supervision. They did what MDRC and most scientists do, which is an observational study, meaning they took a lot of statistics from the past and then they tried to associate them with results one way or the other. Great thing to do as a first step. not anything close to the scientific rigor that a randomized study such as the one that we conducted here did. And so we are suggesting that the county consider a randomized study because we hope, put it this way, that this one, although it took a long time, as these studies do, produced useful findings. And these are as close to scientifically unassailable as one can get. There is no platinum standard to methodology randomized as the gold standard.

The questions.

I have a question, Judge. So two things. You said that cash bond or bail does not work.

for all but a very, very small fraction.

So that was the second part of my question. So take that one first. What was that very, very small fraction? And did you identify those folks for Dane?

The answer is that at present, we do not have the statistical techniques to identify those folks, who they are at the time of the first appearance decision. That is another research project that we have recently gotten seed funding to undertake. And none of the risk assessment instruments that are currently on the market or in use anywhere attempt to do that. And so that's, again, the biggest point for us, which is why it's a research project. No one is asking that question. There's an author whose prose I despise named Thomas Pynchon, but who nevertheless had for me a great line. If they have you asking the wrong questions, they don't have to worry about the answers. And right now, all of the risk assessment instruments across the country are asking the wrong question. They are asking, what is the risk that a person will misbehave if released? they are not attempting to answer the question, what is the fraction of risk reduction that comes from a cash bail-in position? Those are two entirely different questions. One says, how risky is it if somebody is released? The other one says, can you do anything about it with cash bail? And our analysis proves pretty firmly that for the overwhelming majority of people, you can't do much about it with cash bail. There's only a very small fraction. Now, it is a policy decision to say whether the imposition of cash bail for I'm making this up 30% of people, which is roughly where we are in Dane, right, is worth it to catch 3.5% of people whose behavior can be changed if you have the 3.5% within that 30, because some of them may have escaped to the 70, if you're taking my meaning, right? That's a total policy decision. It depends on the consequences of the new violent criminal activity, right? Nevertheless, if we could find those and identify with any kind of precision, those 3.5%, that would be a vast advance. That's what we have seed funding right now to attempt to do. But my big picture most important point is right now none of the risk assessment instruments in use try to do that. So again, they're all asking the wrong question. So this might be a wrong question too.

Okay. So we know there are three states that don't have cash bail. Wisconsin is not one of those three. Have you seen jurisdictions where decisions at the initial appearance, I'll give an example and then tell me if this occurs anywhere. Somebody comes, they may have had two FTAs, they got a signature bond on the first, The second comes, and the recommendation from the prosecutor is $250. Defense counsel might say, this person doesn't have any means. We would like $25. The commissioner weighs that and then makes a decision. Have you seen in any jurisdictions that cash bond is part of their policy? work that various jurisdictions have changed into sliding scales or making a different decision from a county perspective versus a state

We have done none of this work ourselves because we only do randomized studies. Others have done work in which they have studied, for example, the before and after. They've compared before and after, and this is what I think you're talking about, tell me if it's not, a legal intervention that mandates that commissioners or magistrates consider the means of the individual. when setting bail. And there are a few of those studies, however those policy interventions are relatively recent. And so I don't trust yet the results. And because I don't trust them, I don't know them. And I would counsel against over-reliance on them anyway because many of those changes are recent. Where, again, what has happened is the state legislature in something that perhaps a constitutional amendment has come in and said, when setting bail magistrates, commissioners, you shall now consider the means of the individual. And I think that's what you're asking, right? Is the commissioner now taking into account the fact that a $25 cash bail-in position is the same thing as a remand, right, as an incarceration?

Yeah, I mean, overall. But I was kind of thinking, have you seen that difference between kind of state constitution and then in practice in various jurisdictions that they would then... have a potentially a different scale than was envisioned.

I have not seen that. I will tell you that what I do know is that changing the law doesn't necessarily change commissioner magistrate decisions. That the latter turns out to be extremely difficult to do. They are the commissioner magistrates are typically more responsive to things other than changes in legislative or constitutional. That's across the country. I would not expect it to be true here, but it is true across the country.

question that's right along with that. So have you filtered out the difference between cash bail set as ordered versus cash bail posted? Because obviously there's a huge difference.

No, that we are studying the moment of the decision of the commissioner. the commissioner doesn't know whether the cash bail will be posted or not posted. And so the idea is that's part of, to analogize, that is part of the mechanism, it's part of the, if you give somebody an antibiotic, do you wanna separate the bacteria that it kills versus the bacteria that it doesn't kill? And the answer is no, what you wanna know is the patient gets better, right? And so the idea is, well, some people, some bacteria die, some don't, is it enough? Right. In other words, you don't want to separate out what happens too far downstream from the thing that you're studying.

I'm just adding the variable of the amount of cash bill making a difference.

Yes, we have done that. We have separated out the amount of cash bail that the commissioner imposes. The analysis is the same. So we have simplified this down to categories. Our previous analysis two years ago distinguished between low and high, and that didn't matter. And since we were going to hit you with enough concepts and reverse a designation on a chart, which I'm eternally embarrassed about, we decided that we would just simplify everything down to just yes, no, right?

Other questions?

So if we're looking at this chart right now, we were certainly able to identify from the study that the assessment from or to the commissioner didn't really have a difference on these overall impacts. But the overall impacts themselves, would we be saying that 3.5% of

Sorry.

Violent crime is preventable?

No. Oh, I see. No. 3.5% of people are in the preventable category. So the- Not necessarily we would capture that group when we're- Well, on top of that, it would not be 3.5% of violent crime that would be preventable because the amount of violent crime is very low. So if you caught only 3.5% of it, we wouldn't even be able to tell, right? So if you could perfectly identify the 3.5% of people who were preventable on violence, you would catch more than 3.5% of the violent crime. Because only about a total of around 8% or 9% of people commit violent crime at all.

So then taking that back to that next category, so just the crime at all. So we have a higher 11% of that. We have a higher rate of lesser crimes. These are not crime rates.

These are people. Right, right, right. So the rates that you're talking about are in some sense represented in this graph, although they're not raw rates in this graph. Sorry. They're not raw rates because, again, we had to separate out, the math told us that we had to separate out the first arrest from the second arrest from the third arrest within the study period. And so these numbers at the top are kind of statistically fuzzy, right? It's not like you have a 30% new criminal activity rate. We have that in our previous report from two years ago. We can give it to you, or you can just find it in that report from two years ago. We have just the raw rate. For comparison purposes, that isn't the number you want, but we have just the raw rate. It's around what? Yeah. 30% anyway.

Okay. It's not too different from that. Right.

It's not far different from it. So it's around a 30% NCA rate here in the county, which is higher. And I think there are two reasons for that. One is, again, the amount of cite and release that y'all do means that you have a primarily 70% felony population as opposed to 70% misdemeanor. And the second is because of the amount of signature that you use. Mm-hmm.

Last couple of questions. Yeah.

So I don't know if this is a little too into the weeds, but at one point you talked about how the increase, like the moderate effect on the use of cash bail or the decisions made by the commissioners, that it was like a 2.5% increase in cash bail. Cash bail then would lead to incarceration to some degree for however long. So I know you talked about this very briefly, but I'm just wondering your thoughts in the introduction, you talked about how there's a lack of effect then on the predisposition and incarceration days. What are your thoughts on, I guess, that going up small, but statistically significant?

We think that the primary mechanisms are people actually posting the cash bail and subsequent modifications of the cash bail decision. neither of which we have any, I mean, again, we hope that as little as possible as this has been our own value judgments. I will just say we have no problem with either one of those things. In addition, in other jurisdictions, not in Dane, because in Dane you don't have a bail bonds industry, but in other jurisdictions, the laws of the bail bonds industry are so intricate that the imposition of cash bail doesn't actually necessarily mean the person will actually ever post cash bail, even in order to walk free. Nor does it mean, by the way, that if the person fails to appear or commits new criminal activity or whatever, there will ever be a forfeiture event. In some jurisdictions, the fraction of forfeiture is so low as to make the quote-unquote deterrent effect of imposition of a bond a bit of a joke.

Anything else? Yeah, go ahead. I know that obviously you work with these other jurisdictions. Have they made any changes to their usage of the PSA?

Not yet. These are relatively recent results, and we haven't done a full report. In Polk, the state legislature, at the inducement of the bail bonds industry, legislated the PSA out of existence. So they banned the PSA because the... The bail bonds industry thought that the risk assessment instrument would result in a vast decrease in the imposition of bail bonds. Our results here, for example, suggest that was comically untrue, that in fact the opposite occurred. And in fact, I think that happened in Polk as well. Do you remember? No.

Right, and Polk did impose more OR.

Oh, that's right. They did at the lower incident scale, and overall, they did overall.

Yeah, that's right. But it was nowhere near, it didn't kill their industry, no. Yeah, I will say, ironically, for the bail bonds industry, while it did result in more imposition of own cognizance release, it did also increase the average bail amount, so they would have probably broke you. Right.

Each day each one that was imposed got more expensive. That's what I was trying to remember so they impose fewer of them, but they imposed higher amounts. And so they probably would have seen made the same amount of money.

I'd like you to just stay there, but the speaker that was going to do the Richard speak before it has joined Clark Rogers and we might have to see online and we can switch it.

And some questions after OK.

Clark, I think you're free to talk. We may not see you, but go ahead.

We're not hearing you, Clark. They may be trying to promote him, so he'll have to be promoted.

Okay, Clark, are you there? Okay. Well, we'll hold up.

It says that he is being promoted and he has been switched over to the panelists right now.

Okay. Clark, if you can hear me, speak up when you're ready. Okay.

He is unmuted and he has been promoted to panelists.

Thank you. Clark, you're free to speak if you're there. Okay. We'll try one more time later, but I know there was at least a few more questions here just in the interest of time of DAO's end.

So it's good to see your findings because I think our fear was if we had the PSA or not having a PSA, we might end up guess incarcerating more people at the beginning of the process and I think Commissioner Moyer when he was here is what sort of really sparked this relationship and concept of should we look at what we do and how we do it the supervision did you find what did it have the most impact on failure to appears new crime and violent crime all the above

We have done virtually no work at all on supervision. That's our recommendation for that if the county is interested that we together study supervision rigorously. We just don't have anything for you on supervision.

And for the most part, this is looking just at people who are released, right? I mean, if there's a cash bail decision that causes someone not to be released, they don't fall into your debt, or do they?

They do, and that's what we spent the last several years pulling in order to be able to answer that question. Again, this is a different form of his honors question earlier. is, for example, as I mentioned, this graph is subject to exactly the problem that you just articulated. This graph is not. This is, again, the advances in the statistical technique that depend on the assumption that we have all the information the commissioner has. And so if we have that information, again, that was true in Utah. We had all the information that the magistrate, as they call them in Utah, that's true in these graphs at the bottom are sort of rock solid, right? But they were exactly to answer that question. And so they do not depend on that problem. They are unaffected by that problem. Okay. So because the change now is we get to argue dangerousness. Right. Which we did not when we started this process. That change, that was a constitutional amendment, if I recall correctly. That change had not occurred. It occurred after the end of 2019. Right. And that's correct. Right. So for what it's worth, they can argue the legal framework in both Polk and Utah allow the prosecution to argue, well, sorry, let me be careful. They allow the decision maker to take into account dangerousness in both Polk and Utah. Now, it's still not exactly identical to Dane because in Polk, there is no defense attorney nor prosecutor present. at the in-custody first appearance decision. And in Utah, there's no hearing. The magistrates are making these decisions on buses. When somebody else is driving them home in a car, I hope they're not doing it while they're driving a car, but you get the idea. They're making them just sort of whenever they can. And they're all on computer files on the net. So there's no argument. There is a colloquy in Polk, but no attorneys. And there's no hearing in Utah.

Are either of those two, Utah or Polk, grand jury models?

There are grand jury systems available in both states. They are rarely used. And none of these cases that went into our study came from a grand jury indictment. Thank you.

It's really interesting. Thank you for your presentation. You mentioned that changing the laws doesn't necessarily affect the commissioner-level decision on whether to impose cash bail or not. What, in other jurisdictions, has?

Changes in culture. And there's a whole literature, a whole field, a whole set of scholars that study how to change culture. It turns out to be extremely hard. There are some instances in which bail decisions change as a result of sort of word from above and within the judiciary. Sometimes that works. I think based on the people that I've met here in Dane, that there would be changes in response to the law as written. But this may be the only place in the country where I'd say that.

Is there, can I just do a quick follow up? Have you seen other jurisdictions provide more like commissioner level data, like publishing, like rates or decisions or like- I'm going based on the findings of other scholars, supervisor. What's that?

The other scholars have done these studies and they have gotten that data. They haven't typically published it. So we haven't done any of this analysis ourselves. But the other scholars have gone in and worked with those jurisdictions. Some of the scholars are very good. So I trust their findings enough to try to relate them to you. And so we haven't done any of that work. But the other scholars go in and get the data and look at before and after. Now saying, when you compare before and after, you don't know whether for sure the change caused the effect, right? Because the change could have been just coincidental with things that were happening anyway. So it's a little bit iffy. But it turns out we think we do know, we here as the scholarly community, we think we do know that it is extraordinarily difficult to change culture and that culture eats the law as written for lunch. In other words, if you change the culture, it doesn't matter what the law is written as.

Jeff, did you have a question?

Yeah, just quickly. I was hoping that you could go a little bit deeper into what you mean by culture. Are we talking about the culture of the judiciary, the culture of policing in a specific community? What does that look like?

I was referring to the culture of the decision makers, in this case, the commissioners here in the county. Most jurisdictions have a larger number of individuals making these decisions, either because there's more volume or because they rotate them. a they they rotate them in and out commissioner or if I correct I'm correct you were doing this for something like 2 and a half decades for for what only 30 years I'm sorry 3 decades I was off by half a decade and you were the primary person doing that for those 3 decades.

Probably 80 to 90%.

And so and I believe for the majority of the time that during our study period. Commissioner Hansen was the primary decision maker from sort of 20. So culture here would be kind of a small thing. It would be kind of one person, habits of the decision maker.

So then have you seen an impact on how an aggregate that affects how these systems are restructured?

I have, I haven't, I don't know the answer. So it's not that I have seen the absence of it. In other words, it's not that I can say I've looked for it and it's not there. It's that I don't know the answer. I'm the wrong kind of scientist for it. Thank you.

We're going to try one more time.

Yeah.

And which one? I think attendees. No. Clark, did that make a difference? Sharron, I lost your phone.

Oh, there it is. Allow to change to attendee or what? Right now he is a panelist.

It's Clark saying on your phone. We're trying to get the public speaker. So he should be unmuted. Clark, now.

I'm here. Can you guys hear me? All right. Sorry. I just wanted to be my my, my sole issue was just to be be involved if they if anybody had any questions about how the public safety assessment was worked over the last 10 years. Dr. Greiner, I think, covered a lot of that in one of the early slides where he discussed the flow of how everything worked. And that was kind of where I was going to start at because when Rhonda, Frank, Lauren, and I first initiated the actions of what led to the data points for the study, Rhonda and I worked on that and worked with the A2J people and with and with colleen to make sure we were following what the study protocols were were for and we have always tried to uh the my favorite phrase is we we try to maintain fidelity to the instrument and i think from from my point of view a lot of a lot of my concern has been as long as we're maintaining fidelity of the instrument we're showing what this is supposed to do But then part of the problem becomes it's not just the instrument that's being seen by everyone because you have decisions by the commissioners, but you also have decisions by the district attorney's office and by the public defender's office. And they all have their own perspective of what should happen in those cases. So it's like the decisions that happen I think are not really done in a vacuum. And my concern has always been, is it the study that's the problem or is it the application? So that was my concern.

Thank you. Clark, it's a pleasure to hear from you again, by the way. I'm sorry. It wasn't until I heard your voice that I remembered I recognized you. I'm sorry. I'd forgotten who Clark Rogers was. But now, of course, I remember. It's good. It's a pleasure to hear from you again. Thanks, Jim. All I will say on this, and we even had slides, as you can see, most things that we try as human beings don't work. The failure rate of new cancer drugs that make it to expensive testing is 90%. So 9 out of 10 drugs that drug companies think are worth investing into, sinking millions of dollars into testing, never make it to market because they don't work. And there is no doubt, absolutely no doubt whatsoever, that the PSA was produced here better, and to use your term, Clark, with more fidelity than here than it was produced anywhere else in the country. The problem is the instrument and the state of our scientific knowledge. And so it should not surprise us that it doesn't work. I think a lot of the things we do currently as human beings don't work either. And again, that's sort of our point about cash bail is for the most part, it doesn't work. It's very, very hard to sign things that work. So I don't think that means that we stop trying. There's just too much, our problems are too large to say that this is the best we can do and give up.

Well, our relationship with you certainly worked best for some. Thank you.

Real quick, finding things that work and changing behavior.

All right, so cash bail has little or no effect, right? But going off your analogy about drug treatment, for me, I get migraines. I can take my migraine medicine by itself. Sometimes it doesn't have any effect. I combine it with an Aleve pill, and I'm almost guaranteed that it will work. So you talked about supervision. Are there other things that we can be including in that, almost services or case management kind of things that can impact supervision? behavior.

We have no strong gold standard evidence yet that anything at a bail decision for a custodial first appearance decision does work. We have not studied supervision. No one has rigorously studied supervision or services. And that's why we suggest that that's something that the county consider and that we would love to be involved in if this relationship hasn't been too unpleasant. We propose to do that to see if supervision and or services or in combination. The reason why we thought it was worth studying bail alone was because of the high costs of incarceration, that that is sort of the costs associated with that of incarceration, meaning bail, thus some incarceration are sufficiently high to where if that is not working, then it may be that it's time to take a serious look at it because the costs are so high on the county, on the jurisdiction and on the individuals and on the families and on the fabric of the community. So we would love to be involved in exactly the question that you ask, Chair. Thank you.

Okay. Again, thank you very much.

Thank you.

Our pleasure. So we're going to wrap up the meeting, but so we don't have reports for the pre-trail services meeting. We've got some future dates.

And then just adjourn probably. Okay.

The free trial we do most of what we do so move and their second second second and we will adjourn the formal for us or the subcommittee. And we're jerk.