Probable Causation

View Original

Episode 79: Jenny Williams

Jenny Williams

Jenny Williams is a Professor of Economics at the University of Melbourne.

Date: August 30, 2022

A transcript of this episode is available here.


See this content in the original post

Episode Details:

In this episode, we discuss Prof. Williams's work on electronic monitoring as an alternative to prison:

“Can Electronic Monitoring Reduce Reoffending?” by Jenny Williams and Don Weatherburn.


OTHER RESEARCH WE DISCUSS IN THIS EPISODE:


TRANSCRIPT OF THIS EPISODE:

Jen [00:00:08] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an economics professor and the director of the Justice Tech Lab. My guest this week is Jenny Williams. Jenny is a professor of economics at the University of Melbourne. Jenny, welcome to the show.

 

Jenny [00:00:26] Hi Jen, thanks for having me and so excited to be here.

 

Jen [00:00:29] Today we're going to talk about your research on electronic monitoring as an alternative to prison. But before we get into that, could you tell us about your research expertize and how you became interested in this topic?

 

Jenny [00:00:40] Oh, well, I have worked my whole career, which is quite long, because I won't tell you when I graduated a long time ago. You can look me up if you really like, but I work in the area of the economics of risky behaviors and so mostly substance use and crime, so I like to study people's foibles. I think that's what makes them most interesting and so my expertize is in using data on individuals, which could be either administrative data or survey data to conduct causal analysis to better understand individual's choices and then how their choice is going to be shaped by policies.

 

Jenny [00:01:15] And what's so interesting about doing research on crime and substance use is that a lot of people, including policymakers, have really strong opinions about policies related to these behaviors, and they're not necessarily grounded in evidence. So there's there's a lot of opportunity to contribute to building a sort of robust causal evidence base to hopefully enable policymakers to make better informed choices and improve policy design. And in terms of working in the economics of crime, I've been interested in this area since graduate school, and my dissertation was actually sought to build on Becker and Alex models of crime by including social capital, because I could see how useful Becker's model was about creating levers to alter people's behavior by changing the relative price of behaviors, but I felt that there was things left out of the model that are also very important for determining people's choices.

 

Jenny [00:02:10] And so I drew on the social capital literature from Criminology Lab and Samson in particular, and Jonathan Coleman. So of course my dissertation, the work I did for my dissertation predates the availability of administrative data, and the data I used was from the 1958 Philadelphia birth cohort study. That was just the most amazing collection of data of individuals everyone born in Philadelphia in 1958, and it contained self-reports on criminal behavior. So that's sort of what I started in crime in grad school, and that's what my thesis was on, but after grad school, I kind of moved from crime to drugs and became very interested in decisions regarding drug use and policies around drug use. But I've always sort of dabbled a bit in crime, and that's sort of how I came across Don Weatherburn, who's a super interesting guy who until recently led the Bureau of Crime Statistics in New South Wales in Australia, and we stayed in contact over various things and at one point he wrote to me and asked for my advice about something he was looking at, and that was the work on electronic monitoring. And so when he asked me a question an econometrics question, I said, Oh send me the data Don let me have a look and we just went from there. And look, Don Weatherburn has just the most excellent nose for important policy relevant questions and the data to answer them when it comes to criminal justice in Australia. And his data that the Bureau of Crime Statistics that he was in charge of collecting and providing is just exemplary. It's just really, really good quality.

 

Jenny [00:03:44] And with regards to the question of electronic monitoring, he really picked a really good question because prison populations are increasing a lot in New South Wales where he's based and electronic monitoring has been used for over 30 years all over the world and it's really a credible alternative for imprisonment. And, well, there's a lot of evidence questioning that idea of whether imprisoning offenders at the scale that's currently done reduces criminal behavior. And so if sending offenders to prison doesn't reduce the likelihood they re-offend, and given the high cost of imprisoning offenders and their families and the communities that they live in, it's really worth considering alternatives. And as far as electronic monitoring goes, it's been used a lot. So it's natural to wonder whether it might be, you know, a viable alternative to incarceration, but there's really not much causal evidence around it. And so it just seemed like a ripe, low hanging kind of fruit project, given the data that Don had at his disposal through the Bureau of Crime Statistics and his knowledge of the criminal justice system, it just seemed like a really good project to become involved in.

 

Jen [00:04:53] Yeah, that's fantastic and I totally agree. There is shockingly little evidence on this, which is why I was so excited to see your paper. So your paper, the one we're talking about today, is titled "Can Electronic Monitoring Reduce Reoffending" as you said, it's coauthored with Don Weatherburn. It's also been published in the Review of Economics and Statistics and in this paper, you consider how sentencing nonviolent offenders to electronic monitoring instead of prison affects their likelihood of being charged with new offenses in the future. So let's start with some background. What is electronic monitoring and how is it typically used in the criminal justice system?

 

Jenny [00:05:29] Well, electronic monitoring is also known as tagging. So electronic monitoring is a practice of putting an electronic transmitter physically, like a bracelet worn on an ankle or wrist on an offender to monitor their location remotely. Sometimes that's outsourced to private companies that and the criminal justice sector can rent the actual devices or sometimes they buy them and do the monitoring themselves. So the idea is that offenders are typically required to stay in a specific location, typically their home, and they are monitored to ensure that they do so with an electronic device. So it's typically used in conjunction with a requirement to stay at home, although the conditions of electronic monitoring can include attending a preagreed activity such as work or attending doctor's appointments, but pretty much they have to stay at home under electronic monitoring. But there are so that sort of if it's used as an alternative to home detention, it's also used as a form of early release, as a condition of bail or for pretrial detention.

 

Jenny [00:06:35] So it doesn't sometimes it can just be used in conjunction with like a curfew. So people just have to be at home, for example, between 7 a.m. and 7 p.m.. So it's just a way of monitoring individuals to ensure that they comply with these the requirements. And it means that they because they're staying at home, it's not costing the public to house them, feed them the clothes and and the supervision costs are much lower, but it does reduce their liberty when it's used as an alternative to incarceration. It's sort of this a lot of checking and other requirements put on the offenders and so sort of like minimum security prison with day release, for example, is how to think of it.

 

Jen [00:07:21] Got it. And then and then how is electronic monitoring used in New South Wales specifically? And I guess also how long has it been used there?

 

Jenny [00:07:31] Oh, well, the program that I evaluates been available in New South Wales since 1997 and it's designed to be an alternative to prison for less serious nonviolent offenders and the goals of the program are to divert people from the prison system and to rehabilitate them. So as with other electronic monitoring programs, this program that Don and I evaluate involves offenders serving their sentence in the community rather than in prison, but with restricted freedom of movement. So offenders yet they live at home either alone or with their family, and they're not permitted to leave the home, except when approval is given by the supervising community officer for activities like work, education, intervention programs, drug treatment, for example, or medical appointments and the activities are scheduled on a weekly basis, but the compliance is really tight. It's considered a breach of their conditions if they're late home, if there's a train strike, it's really, really tight. Well, it's investigated as a breach, and then a committee will decide whether it's actually considered a breach.

 

Jenny [00:08:35] So their movements, to the extent that they can leave the home, is tightly restricted and monitored. They basically have to be at home. They call throughout the night to check that they're there and so if women who have babies or small children, it was a nightmare because you get your baby to sleep and they get a telephone call because the program I evaluate is for, you know, GPS tracking. So they were there was just a device to check that they came into their home and when they went in and out, it kind of beep to let the monitoring company know. People think all these people are staying at home. They can drink, they can party, they can hang out with their friends. No, they're not allowed to drink or take drugs. They're subject to drug testing. So it's very restrictive if they're out at work, there's visits by filled up out in the field by their supervising communities correction officers.

 

Jenny [00:09:23] So their behavior is very tightly supervised. And as I mentioned, the program also has a rehabilitation component and so all aspects of the offenders lives are taken into consideration in designing this the offenders specific programs and so they take into consideration family issues, family parental responsibilities, drug and alcohol use. So offenders might be expected to be employed or undertake vocational training. They're required to undertake counseling or other programs to address personal issues of which is often drug and alcohol related. And importantly, offenders serving sentences under electronic monitoring are supported in rehabilitation by a high rate of face to face contact with supervising officers. And those officers liaise closely with supporting community agencies.

 

Jen [00:10:15] It's interesting, as I think we're used to thinking of the main alternatives. If you're sentenced as being prison or something like probation or community supervision, where there's, you know, you don't have this like GPS monitor or electronic monitoring, but you might have all the same meetings required. In theory, what you're describing you could imagine having in a community supervision or probation program. I think of electronic monitoring basically being in between those two things. It's like probation, but you have your liberty is much more restricted. And they really keep track of where you are. Is that the way to think about it?

 

Jenny [00:10:46] Well, that's true. So if you don't comply with the requirement to stay, well, they actually have to stay at home. They don't have liberty. I think that if you those other programs that you mentioned, like probation, I think they have freedom of movement.

 

Jen [00:11:02] They do, yes.

 

Jenny [00:11:04] So I think that's a huge distinction. And when you read, you know, about people who talk about their own experience, they find it very challenging to stay at home and be restricted in that way.

 

Jenny [00:11:15] So and I think that we've all experienced that.

 

Jen [00:11:17] Yeah.

 

Jenny [00:11:18] It's, it's it's it is, it is quite challenging. And if you don't comply, they're easily detected. Also with electronic monitoring, the supervisor, there's an awful lot of face to face interaction with a supervisor who supports them and I think it's much more of a supportive kind of arrangement. And if they breach, if they fail a drug test, the chances are, rather than write it up as a breach, which would have them sent back to prison because they've breached the term of the agreement, they could be placed in in-patient drug rehabilitation to assist them and support them. So I kind of think of it as it's rehabilitation that's very supported and so if they fall over or face challenges, which they're going to do, there's people who can help them navigate those challenges.

 

Jen [00:12:05] Okay, interesting. So as people are probably getting a sense for, there are several dimensions to this program. So so let's just go through what the various ways are that electronic monitoring is different from prison, which is the alternative you're going to be comparing it to here. In other words, what mechanisms should we have in mind for why an electronic monitoring sentence might affect someone's outcomes?

 

Jenny [00:12:28] So the big difference, I think, between electronic monitoring in prison is that in prison you're removed from the community. And you're all you have no control over your movements or or your diet, you have no autonomy, no liberty, no freedom. And that is unpleasant, as we've all experienced under COVID. So we know that that's a huge deterrent effect there. So prison has a huge deterrent effect on re-offending, whereas electronic monitoring, you're at least you're in your home, it has a less deterrent effect.

 

Jenny [00:13:03] And so the deterrent effect is specific to saying, I've experienced this, I don't want to experience this in the future. However, while prison does have this mechanism, this strong, specific deterrence effect that will reduce re-offending. It turns out in the literature that that's not all that's going on, because it's not necessarily clear to the extent to which that effect dominates other mechanisms and those other mechanisms, things like, well, when you are in prison, you're hanging out with a bunch of other people who have expertize in crime, so you're able to build your criminal networks and your criminal capital. When you're home detention, you're not permitted to have those kind of interactions those actions are prohibited. So you are not able to build your criminal capital so it's less criminogenic from that, just being placed in prison allows you to be better at committing crime, which means you might be better at evading crime also.

 

Jenny [00:13:59] But with home detention or electronic monitoring, while your freedom is restricted, you're kept living within the community and not only are you not interacting with other criminals, you have programs which involve work or employment. So if you aren't able to find a job, you are required to undertake training programs to increase your employability and build your skills and connections to more legitimate institutions and you live with your family, maintain those connections. You are required to address other things that might contribute to your offending. It might be anger management, drugs, substance use, and all of these these forces that act to restrain your offending, you know, a balancing of balancing out perhaps a smaller specific deterrent. So the question of whether or why electronic monitoring might lead to lower offending than prisons boils down to the question of the relative size of the net effect of specific deterrence and factors related to criminal and legitimate networks and ties that you really build and develop under electronic monitoring, which you're not able to do.

 

Jenny [00:15:14] In fact, they're weakened and broken when you're sent to prison. But it's an empirical question as to what dominates specific deterrence or these positive network and social ties that you develop under EM and that's what the paper seeks to address. We can't kind of pass out the different effects. We can only measure the net effect, and that's what we're doing.

 

Jen [00:15:36] Yeah. I also found it really interesting in the paper you mentioned that. So there are obviously a lot of these rehabilitative programs, drug treatment or so on, that one might that might be included in someone's electronic monitoring sentence. Those programs are also available at most prisons, but there's no incentive or there's no we require people to engage in that programing while they're incarcerated, but there's either you know, if you're on electronic monitoring, I guess you either engage in it or you go to prison. So there's a real incentive to actually do the things you're supposed to do or just something else about the program that they're actually able to require this this engagement, these really rehabilitative programs and that also just seemed like a really useful distinction here. That is sort of another reason to be interested in these programs.

 

Jenny [00:16:26] Yes, it's it's true. In prisons in New South Wales, if your sentence is greater than six months as long you have available rehabilitative programs, but they're voluntary. And I think another issue is that they're not in the world that you're going to have to live in post prison. And so once you're out, if you encounter problems, yes if you're rehearsed or practice skills, they're available for you. But if they don't work or you forget and panic or you make a mistake with electronic monitoring, you know, you actually have people there to help you navigate the real world when you're actually facing these problems.

 

Jenny [00:17:06] And I think that's a bit of a difference that doing things in theory and practice, as anyone who's ever done any applied econometrics research would know you know, in theory, you have an instrument in practice, it might be theoretically a great instrument, but still might not work in practice. So there's a difference between theory and practice and I think that that two distinctions, one, between rehabilitation in prison and under electronic monitoring, is that it's something that you can volunteer, you can choose or elect to do in prison. With electronic monitoring it's a condition that you have to agree to with electronic monitoring and if you don't comply, then it's a breach and you can be returned to prison. And the second is, if we're learning something in theory is different from doing it and learning and experiencing it in practice with support and I think that that's another benefit of electronic monitoring.

 

Jen [00:18:01] Yeah, it's so interesting. Okay. So we mentioned at the outset that there isn't much research on this, but there are some. So for your paper, what had we previously known about the effects of electronic monitoring?

 

Jenny [00:18:13] Well, I'm only aware of three papers that have studied or compared electronic monitoring to prison in terms of impacts on reoffending. And the first was set in Argentina by Di Tella and Schargrodsky I and probably get all these names wrong. That study looked at pretrial use of electronic monitoring in a very extreme situation where the prisons are really awful and the criminal justice system doesn't work terribly well and also, it was available to anyone, people who raped or murdered it was so it wasn't it wasn't targeted and it was in pretrial, but in a very extreme circumstance, which isn't very similar to Western countries, I wouldn't think.

 

Jenny [00:19:00] But it's a so it was such an exciting study, a such novel, innovative study, but as I mentioned, it was available for all crimes, this program. But they found that electronic monitoring reduced recidivism by between 11 and 23 percentage points relative to prison, so they found it was effective even in this very extreme setting. And then so that's at a pretrial using electronic monitoring before people come to trial, so they've just been charged with a crime. And now there's Olivier Marie, who evaluates the UK's use of electronic monitoring and its used as what's referred to as back end or early release alternative and in that setting, prisoners have to agree to a curfew, so they have to be at home between 7 p.m. and 7 a.m. so it's a bit different because people have to stay at home all the time. But even so, they found it reduced recidivism by about six percentage points, relates to imprisonment after one year on electronic monitoring.

 

Jenny [00:20:02] Now there's another study which evaluates piloting of electronic monitoring in France, and it's the most similar to the context I study because it's a front end alternative and by front in, I mean it's an alternative to imprisonment. So they rather than spending your sentence in prison, you spend it under electronic monitoring in your own home. So they evaluated the French program and once again, it allowed weapons charges and sexual assault. So it's slightly different to my setting because it did allow these violent offenses and I point this out because when I think of evaluating programs, I want the programs to be relevant to policymakers and I imagine trying to talk to a policymaker and convince them of something and so I was a bit surprised that it included weapons in sexual charges in who qualify for this program.

 

Jenny [00:20:52] But they in the French setting, electronic monitoring was found to reduce recidivism by 6 to 7 percentage points relative to prison after five years. So that's quite a long term impact. And so that percentage point change translates to 11% reduction in recidivism. So that's have they ever offended after five years so all the studies that I'm aware of, those three that occurred before my study, they all found that electronic monitoring reduced reoffending compared to incarceration in prisons.

 

Jen [00:21:24] Yes. So why didn't we have more than three studies? What do you think of as being the main challenges or hurdles for researchers like yourself when approaching this question, as has it's mostly been a data issue or mostly an identification issue, or is it both?

 

Jenny [00:21:39] Well, I have to say, data, you have to be able to have data and so well, both data and identification. So prior to the advent of these beautiful administrative data that have become increasingly available, I mean, it would be very hard to evaluate these programs using survey data. So you do need high quality individual linked data and that's why the Argentinean study was so impressive. They went in there and actually transcribe handwriting records, actually went in there and collated all that information. That was that was just stunning. The subsequent papers have had access to administrative data and in day man was doing an evaluation, but in addition to having access to administrative data, the challenge is identifying the causal impact. Because when people when judges are selecting candidates for electronic monitoring, they're choosing people who they think are a low risk of re-offending because they don't want to be on the front cover of newspaper saying. They released someone who's now gone and been, you know, mass murdered people.

 

Jenny [00:22:41] So they really looking to select candidates who they believe are low risk of re-offending. So if you just compare the offending rates for the group who served a sentence under electronic monitoring and the group that served their sentence under prison, you, of course, expect to find lower re-offending among electronic monitoring people because they were chosen because the judge observed something about them, extenuating circumstances surrounding their offending, or their remorse something about them that led the judge to believe they'd be less likely to re-offend. So you can't just compare re-offending among those who serve the EM sentence, electronic monitoring sentences and those who send those who serve prison sentences. But another kind of thing that's a bit tricky is when you talk about evaluating electronic monitoring, electronic monitoring is used in a lot of different ways.

 

Jenny [00:23:31] So you have to be really clear about how it's being used because it could be a condition of bail, early release, pretrial. So it's not a single, well-defined intervention and it's not apply to a well defined set of offenders as I mentioned, it's been used for just anyone in Argentina whereas in Australia, for example, it's just used for nonviolent offenders. So nothing's very homogeneous, but in the setting that I'm looking at, it's used for nonviolent offenders, and I'm looking at it as a alternative, direct alternative to imprisonment because it's a front end. You know, people it's whether you go to prison or you serve that term in your home under electronic monitoring.

 

Jen [00:24:11] And then to measure the causal effect, you're going to use the as if random assignment of cases to judges in New South Wales at the natural experiment. So tell us a bit about how the court system there works and in particular how cases are assigned to judges.

 

Jenny [00:24:27] Okay, so this is where Don was like fabulous, you know, because I always like it doesn't come with the administrative data and so he was really good at learning about this. And if I had questions, he would go to like the chief magistrate and he would ask. So I've got it from the horse's mouth by Don. So the idea is if you're charging it with a crime in New South Wales, your case is going to be heard in a court in the jurisdiction in which you committed the crime. Now, in terms of I'm confining my analysis to Sydney because that's sort of where EM, was mostly available in New South Wales. So in Sydney, Sydney's a huge city. It's the most popular city in Australia, it's the capital of New South Wales it's where the Sydney Opera House is, which you might know, it's a big icon.

 

Jenny [00:25:14] So in Sydney courts are typically multi court complexes, which means there are multiple courts located together and each court has several courtrooms and in every courtroom there is a judge hearing a case. Now, in terms of how cases are allocated to judges, there's a coordinating judge in each complex and they are looking to allocate cases to courtrooms, and they do this to kind of share the workload as evenly as possible. So every week they look to see who's available, who's not on leave, who's, you know, not in the country doing like the circuits for the courts in outside of the city and looking at who's available, the coordinating judge prepares a weekly roster to inform judges of which courts they'll be sitting in in that way, but daily allocations to courtrooms at a multiple complex aren't published till the day before the case is heard.

 

Jenny [00:26:08] So that's no scope for defendants to influence the choice of judge, and there's no scope for judges to pick and choose cases and there's very little notice about in advance of who's going to be sitting where hearing what cases. So these features mean that, you know, within the pool of available judges, the assignment of judge to case within a court complex is effectively random.

 

Jen [00:26:29] Awesome. Yeah. We're always happy when we find situations like this. Okay, so what are the possible case outcomes in the types of cases you're looking at? And who's eligible for sentencing to electronic monitoring?

 

Jenny [00:26:42] Right. So when sentencing a guilty offender, the judge first has to decide whether a custodial sentence is required. So I'm looking at cases where the judge has decided a custodial sentence is required. And if that's the case, then the judge decides the duration of the custodial term that's sort of how sentencing works. Now, electronic monitoring is available as an alternative means of serving a custodial sentence. So an alternative to serving in prison is serving at home if three criteria are met. The first is that the sentence length is no more than 18 months. The second is the offense that the judge is hearing is not related to violent or threatening behavior. And the third condition is the offender's criminal history must not include violent or threatening offenses. So basically, it's available who's eligible well, people who have no history of violent offenses, who are not before the court currently on a violent offenses. And the judge is determined that the offense should be punished with a custodial sentence of no more than 18 months.

 

Jen [00:27:50] And then how exactly are you using that assignment of cases, the judges that we talked about before to measure the causal effect of sentencing someone to electronic monitoring.

 

Jenny [00:27:59] Right. So the way that we disentangle the effect of judges selecting low risk of reoffending, people from the actual causal impact of electronic monitoring is, as you said, exploiting the fact that judges work at courthouses are randomly assigned to cases. But the other thing that's really interesting and the reason that random assignment to judges has any ability to identify a causal impact is that judges differ in their preference over punishments. So judges aren't, you know, like uniform, they're people and they have taste, preferences and beliefs. Some judges don't like to use electronic monitoring. They just won't touch it. They don't believe it's a good punishment for these offenses and other judges do they think it is appropriate for some offenders and so it's this difference. It's the fact that who hears your charge is randomly assigned and that these judges actually just have different tastes over punishment. That's what we're leveraging.

 

Jenny [00:29:06] So, for example, suppose that you and I have committed an offense and we're going to court today and we look pretty much identical. We've got the same offending history and the current charges that we're facing are the same. Your case is heard by a judge A, who is known to not light using electronic monitoring. And my case is heard by Judge B, who doesn't mind using it when they they think it's appropriate. Then, just by virtue of the fact that my case is I'm randomly, randomly assigned a courtroom, and in that courtroom, I find Judge B, I'm more likely than you are to receive an electronic monitoring sentence. This is just due to chance that I happen to be in this courtroom and Judge B is assigned to that courtroom. And it's this randomness in the assignment of electronic monitoring, which results from judges having different tastes and preferences over punishment types and randomness, and who which judge we each see that allows us to identify the impact of i's that randomness.

 

Jenny [00:30:11] It's not to do with your proclivity to re-offend or my likelihood re-offending. It's just who hears the case. So that allows us to separate out whether I'm more likely offend or less likely to offend from the fact that I'm just by luck or a judge who doesn't mind using electronic monitoring. It means I'm more likely to get electronic monitoring, and I can use this to identify the causal effect of electronic monitoring on reoffending amongst those who would have served their sentence in prison know had they faced a judge who was less inclined to use electronic monitoring.

 

Jen [00:30:46] Okay. Let's talk about these amazing data that Don was able to get for you. What data do you have available for all of this?

 

Jenny [00:30:55] Well, actually, I should say that the Bureau Crime Statistics makes these data available to anyone that applies to get ethics approval, but it's you can apply lots of people can use it it's not they make it available to everyone they have produced these beautiful databases that are available to researchers to use. So, yes, the data is super, super neat and we use detailed individual level linked administrative data from courts, prisons and Corrective Services, the New South Wales. Now the database is called the Re-Offending Database or ROD for short, and it was provided by the Bureau of Crime Statistics. And so, as I mentioned, the Re-Offending Database links individual case data from all criminal courts in New South Wales with prison data and then information on judges was not available in the data set when we started this project, but it is now. So for this project I asked Don and he was able to get that information, the judge overseeing each case and then that data was linked to the Re-Offending database, as I mentioned.

 

Jenny [00:32:01] Now that's a standard variable in the approach data set. The second source of data we needed was referrals from New South Wales Community and Corrections Services because before a judge can hand down a sentence of electronic monitoring, the offenders have to be assessed for suitability and that assessment sort of ensures that the offenders are indeed eligible. And so the New South Wales Community Corrections data was provided by Corrections, Research, Evaluation Statistics, unit of Corrections Services, New South Wales, and we were really happy to receive it and it was really good of them to do that, so that was, that was all Don. But so the data has the most serious charge that people are being heard before the courts. So we know the crime type, whether it's the assault, negligent driving, breaking and entering, theft, fraud, all that kind of stuff. We know the offender's gender their age. We also know the number of finalized court appearances in the five years prior to the index court appearance. So we know their offending history, which is really nice. And we also have an indicator for whether the offender had any legal representation during the court case. And we know the local government area where people live.

 

Jen [00:33:16] So what do these cases look like? What are the characteristics of the defendants and their charges and how did those sentenced to electronic monitoring differ on average from those who are sentenced to prison?

 

Jenny [00:33:27] Well, you might be surprised to know the sample that we're looking at is mostly male, 90% male, and their average age is around 31 and the most common offense is theft and traffic offenses, remembering that this program is available for nonviolent offenders. And so about 25% of people in the sample before the courts, with their most serious charge being a theft and a similar amount, about 25% are before the courts they're found guilty of a traffic offense charge and that's most commonly driving without a license. And so they're looking at a prison term or electronic monitoring for these traffic offenses, which I was pretty surprised about.

 

Jen [00:34:18] Yeah, that's a heck of a penalty.

 

Jenny [00:34:21] Yeah. Yeah. So. In terms of differences between the offenders who serve their sentence under electronic monitoring versus prison. Those who end up serving their sentence under electronic monitoring are more likely to have traffic offenses as a more serious charge. So about close to those on EM have a traffic offenses more serious charge compared to about 20% in the prison sample. Those serving an electronic monitoring system are relatively less likely than those serving a prison sentence to have theft as the most serious charge. So about 10% of those of those on EM have theft, a theft charge as their most serious charge compared to about 25% to serve their sentence in prison.

 

Jen [00:35:03] Okay. So we've got differences on average across those going to prison and those going to electronic monitoring. And then are there any differences between those whose cases are heard by the judges that like electronic monitoring versus those that don't?

 

Jenny [00:35:17] Right. So as you've mentioned, they're the people who have electronic monitoring is the way that they serve their sentence versus prison. They have different charges. They're more likely to be having a traffic charge and so that's why we're concerned that if there's differences, on observable characteristics, there might be differences on unobservable characteristics. And it's certainly the case that that's why we are using the randomization to judge who have different tastes for using these different punishments. So really key to what we're doing is trying to ensure that this randomization, which is the basis of, as you mentioned, it's the basis of our instrumental variable.

 

Jenny [00:36:02] You don't want you want to ensure that randomization is genuine. So there should not be differences across more harsh and more lenient judges in terms of the the cases heard. So when we compare offenders in the same court and in the same calendar year, so in other words, controlling on court by year, fixed fixed, we find there isn't any difference in the observed characteristics of cases heard by more lenient and more harsh judges and that's what we hope to find. This is we can't test whether there's differences on unobservable characteristics of the offenders or the cases, but in terms of observable characteristics RTU control for court and and year so you're comparing cases that judges here in a court in the same calendar year. We don't find any differences in observable characteristics in the case it's heard between the more and less lenient judges.

 

Jen [00:37:01] Awesome. Okay. And last data question, what outcomes are you most interested in?

 

Jenny [00:37:05] So the outcomes we're interested in are measures of re-offending. We look at re-offending at the extensive margin, which is just have you re-offended within a certain period of your case being finalized and also re-offending at the intensive margin, which is looking at the number of new offenses that you have before the courts with a certain period after your original case index case was finalized and we evaluate these outcomes six months after the original case is finalized, 12 months, 18 months through to ten years or 120 months after the original case is finalized.

 

Jen [00:37:46] Yeah having that ten year follow up is very impressive, makes for very pretty graphs, I will say.

 

Jenny [00:37:54] Oh, but it really it does speak to the long run effects. And I think that that's really important because you don't want to just be kicking a problem down the road. We want to know whether it whether it's the solution sticks.

 

Jen [00:38:08] Yeah. Yeah. Okay, so let's talk about the results. What do you find is the effect of electronic monitoring on the likelihood and number of new charges?

 

Jenny [00:38:18] This is really exciting because the results show that serving a sentence under electronic monitoring reduces the probability of re-offending from 18 months after the case is heard. It's not significant prior to that event. It also reduces a number of proven charges, and this reduction is evident to ten years after the case is finalized. So, for example, just to give you some numbers to hang your hat from so often in politics in Australia, they look at re-offending after 24 months. So we find that 24 months after case finalization, there's a 21 percentage point reduction in re-offending among those who are sentenced for electronic monitoring. So the reduction is from 45% to 24%.

 

Jen [00:39:06] Amazing.

 

Jenny [00:39:06] Yeah.

 

Jen [00:39:07] It's a big effect.

 

Jenny [00:39:08] Yeah. Then after five years it's a 22 percentage point reduction from 69% among those who serve a prison sentence down to 47%. And by ten years, there's still an 11 percentage point reduction in reoffending among those who serve their sentence under electronic monitoring, but would have served their sentence in prison had they got a different judge. So it's just huge and then in terms of number of new offenses at the ten year mark, we find a 45% reduction in number of proven new offenses ten years after the initial court case was finalized. That's a cumulative number of offenses is reduced by 45%.

 

Jen [00:39:55] The big effects and as you said, this is just luck of the draw. Some people got a lenient judge, a judge that likes EM and some people got a harsh judge. This brings us to a question about compliance. So with analysis like this, the results are relevant to the type of person for whom that randomization matters, right? So for some cases, every judge is going to agree with the outcome should be, but they're going to be some where it matters which judge you get assigned to and then we call those the compliers. So who are the compliers in your analysis?

 

Jenny [00:40:30] Oh, that's a really good question, Jen. The compliers are more likely than the average, often in the sample to be before the courts on a traffic charge, to be female and to not have had legal representation. So I think these results are most clearly relevant to those on traffic charges and those people on traffic charges. The majority of them are driving without a license and or drunk driving, but it's predominantly those driving without a license.

 

Jen [00:40:57] And then the other aspect of an instrumental variable analysis that people might be wondering about is the exclusion restriction. So a key assumption here is that your treatment variable so in this case, a judge's tendency to sentence someone to electronic monitoring is correlated with the outcome only through its effect on the electronic monitoring sentence. That's the exclusion restriction. So since you have random assignment of judges to cases, we're not worried about the usual thing, that judges preferences are correlated somehow with case characteristics, but it's possible that judges might influence the cases they handle in other ways than just the electronic monitoring decision. So this exclusion restriction is impossible to test directly. This is a common challenge in all research, but you dig into the data a bit to consider whether it seems plausible in this setting. So tell us a bit about what you do and what you find.

 

Jenny [00:41:49] Right.

 

Jenny [00:41:50] Yeah, no, I agree with you.

 

Jenny [00:41:52] It's probably buried a little bit in the paper. But I think this is really important because if you think that you're going to have lenient and harsh judges, they might be harsh not just in terms of being more likely to give in prison, but to give longer prison sentences, for example. So if that's the case, then being a harsh judge is also correlated, is also going to be picking up the fact that you're more likely to give a prison sentence in the analysis in control for the length of prison sentence. So the impact we're finding might be that they're actually the instrument might not be valid because it's really picking up stuff to do with sentence length, which has been omitted. And therefore our being a harsh judge is going to be correlated with the submitted variable and cause confounding our instrument won't meet the exclusion restriction because you're really picking up not just the effect of a sentence of electronic monitoring, but also the confounding effect of having a longer sentence.

 

Jenny [00:42:49] So that's my concern actually could be the other way, you know, in that if you get some judges, they might think I'm going in hard and Sharpe going to get seven to prison, but a short sentence and I'm going to scare them straight. And so you have to be really worried about that. Judges are determining not only the manner in which the sentences serve the duration of the sentence, and there's variability in that, too. So the way that I investigate that is that I basically account for sentence length and I treat it as a choice, you know, as an endogenous variable, similar to whether the sentence is served under electronic monitoring in prison. I also model sentence length and I use the random variation generated by the judge that you see similar to the type of sentence, whether it's electronic monitoring, you're randomly assigned to a judge and judges have preferences over sentence length, just as they have preferences over punishment type.

 

Jenny [00:43:52] So I use that random variation generated by random assignment to judge and judges, random variation or judge specific variation in preferences over duration of sentence to directly examine this particular violation of the exclusion restriction and so when I investigate this, I find no evidence that sentence length has any impact on re-offending after accounting for sentence type. So I'm a controlling for both sentence type and sentence length and each are being instrumented. So I'm using this variation generated by random assignment to judges and their preferences over these various dimensions of sentencing. And I just find it really interesting that if once I account for the sentence type, the sentence length doesn't impact on re-offending. I think that raises more questions than it answers.

 

Jen [00:44:54] But it's good for you.

 

Jenny [00:44:56] Yeah.

 

Jen [00:44:57] Yes, but I agree super interesting finding. Okay, so that is all your paper. Have any other papers related to this topic come out since you first started working on the study?

 

Jenny [00:45:08] Well, there's been lots of papers on looking at, you know, whether prison reduces reoffending. But I'm only aware of one additional paper that's comparing prison to electronic monitoring, and that's out of Norway. So there's Norway rolled out an electronic monitoring program between 2008 and 2011 and Andersen and Telle in Norway, I'm probably butchering their names too, they look at how big sentence electronic monitoring impacts reoffending rather than being sentenced with prison. These electronic monitoring is used in Norway for sentences no more than four months, I believe, but they find that electronic monitoring reduced a two year recidivism rate by about 15% and reduced the one year recidivism frequency of offending by about 0.3 offenses on average. So they're just looking at two, one or to two years, but they also concur that electronic monitoring reduces reoffending.

 

Jen [00:46:08] And so what are the policy implications of your results and the other work in this area? What have the policymakers in Sydney taken from it, if you've talked with them and and what would you tell other policymakers and practitioners who are curious about electronic monitoring?

 

Jenny [00:46:24] Well, I think that there's now a well-established evidence that electronic monitoring programs that divert offenders from prison and especially those that incorporate a rehabilitative component, reduces re-offending. Now, you know, whereas France's program was a pilot, this is the program that I study is well-established. It's been operating for like ten years at the point that I'm about, you know, evaluating it. So this is not a pilot program. It's well-established. It's operating at scale for a long duration of time and it's a program that diverts people from prison. It's not a bail condition. It's not early release. So it provides really strong evidence that electronic monitoring is an alternative to imprisonment for nonviolent offenders in particular.

 

Jenny [00:47:17] The second takeaway that I really want to impress upon policymakers is that this reduction in re-offending, it extends out to a ten year time horizon. So the literature now establishes that re-offending is reduced at much longer duration than previously documented and not only does it reduce reoffending at the extensive margin, that's the number of offenses, but a number of offenders, but also reduces re-offending at the intensive margin. That's the number of offenses and the number of offenses is reduced out to ten years also. So the combination of the lower likelihood of reoffending and the number of new charges among those who are placed on electronic monitoring extended over a ten year time period, provides, I think, a really important evidence based on the potential for electronic monitoring to reduce the economic burden of crime.

 

Jen [00:48:11] And am I right in recalling that electronic monitoring is in general much cheaper than sending someone to prison?

 

Jenny [00:48:18] Yeah. Yeah, it is. You know, prisons costs, you know, in Australia, I think it's on the order of $100,000 a year to imprison someone. Electronic monitoring is much, much cheaper.

 

Jen [00:48:31] Yeah. So you get all these benefits, you get all these benefits for less money, basically.

 

Jenny [00:48:35] Yeah. Yeah, I know it's win win. Really.

 

Jen [00:48:38] Yeah.

 

Jenny [00:48:38] It's it's win win. But you have to get buy in from politicians and the judiciary and politicians don't want to be seen to be soft on crime. So even if a policy reduces crime and benefits society, I think you have to be probably quite politically brave to try and explain that to constituents. It takes a lot more words and a lot more concentration from the audience than simply saying tough on crime, tough on crime.

 

Jen [00:49:08] Yeah.

 

Jenny [00:49:08] Which everyone kind of you know, it's like meerkats. Oh, yes, yes. Tough on crime, you know.

 

Jen [00:49:15] Right. Right. It's always easier for, you know, people in the community to point to a news story about one person that was released and committed murder, as you were you were suggesting earlier then to see all the people that are not committing crime now because they did electronic monitoring instead of going to prison. It's just it's the latter is much less salient and harder to to explain to the public.

 

Jenny [00:49:39] It is because I mean, it's people go to prison, come out of prison and commit crimes.

 

Jen [00:49:44] Mm hmm.

 

Jenny [00:49:45] But the point is that if you if they could have been in prison, but they were at home instead and committed a crime. Yes. It's they feel that the policy could have protected the community from that crime by sending them to prison.

 

Jenny [00:49:59] That what we know, what the evidence shows is, in fact. That for the next ten years, from the time of sentencing, the community is going to be experiencing more crime amongst this group of people who would be better off, who should be serving this under electronic monitoring, had they served their sentence under electronic monitoring rather than prison. The concern is that if you if a judge sentences someone to electronic monitoring and they re-offend, it's easy to point to that incident, which isn't very it doesn't agree very often and says that crime could have been avoided, the community could have been protected if that person had been sent to jail. However, what we know from the research is if you look at ten years from the time of sentencing, people who serve their sentence under electronic monitoring commit fewer crimes than had they served their sentence in prison amongst this group who serve this and it's under electronic monitoring just by chance of getting a judge who uses it. So, in fact, the the community saved, I think, about four or five crimes over a ten year period by having the offender serve their sentence under electronic monitoring. And they're not saying that you don't you don't see reporting on crimes that don't occur.

 

Jen [00:51:15] Right. Exactly. Exactly.

 

Jenny [00:51:17] It's very hard to explain to people the benefits of electronic monitoring because it's in terms of things that don't happen.

 

Jen [00:51:24] Yeah.

 

Jenny [00:51:24] Counterfactuals, right.

 

Jen [00:51:25] The counterfactual, right.

 

Jenny [00:51:27] Yeah.

 

Jen [00:51:28] Did you present these results to folks in Sydney at all?

 

Jenny [00:51:31] So I presented it. I presented at a couple of different times to people who work in the criminal justice area, first when it was work in progress, because I wanted to make sure that my understanding of the data was correct to my interpretation of the institutional context was correct before I went beating drums.

 

Jenny [00:51:48] So for example, I questioned to, look, these people are going to prison for traffic offenses, is that right? And yes, I was assured it was. Do they find they don't pay their fines? What else are we to do? And I thought, well, you know, I'm not sure prison is the only option, but for people who have traffic offenses. So I confirmed with folks from the sector that my understanding of the institutions and data was correct. And then at the finalization of the project, when we had written everything up, I presented it to people from criminal justice again to share with them my results, because a lot of them had been so helpful in sharing their wisdom and knowledge of the institutions with me. I wanted to make sure that they saw the results of their input and understood this the the findings from the research.

 

Jen [00:52:38] What was their reaction?

 

Jenny [00:52:40] Oh, they were very polite and nice.

 

Jen [00:52:46] Were they surprised?

 

Jenny [00:52:47] Well, I don't know. I can't they didn't sort of hang around. I guess they weren't academics, so they didn't really hang around to talk. It was a lunchtime seminar, so it was really well attended. I was very grateful the Bureau of Crime Statistics put on the seminar. People came, they listened. They were kind of engaged. They were polite, but it wasn't like an academic seminar. People didn't hang around and chat afterwards.

 

Jen [00:53:10] Got it. Got it. Yeah. And it is this electronic monitoring program still in effect? Has it expanded at all? Is it smaller? What's the current status?

 

Jenny [00:53:21] So what is really, really crushingly just a pointing from a policy perspective is that it's not much used anymore. And that's the result of just a reorganization within the bureaucracy administering the program.

 

Jenny [00:53:37] There was an election where the returning government promised to be to really try work hard to reduce the two year recidivism rate. And the way they were going to do that was by using electronic monitoring, people being released from prison, early release from prison, for example. So what happened was that the electronic monitoring program that is an alternative to prison was collected with the electronic monitoring program for those being released from from prison in terms of administration and as a consequence, you know, it was a one size fits all approach. All offenders were considered high risk and that they had to prove that they should be allowed to have electronic monitoring. So it was a whole shift in the way it was viewed. It became administered in a more kind of punitive kind of way. And they made drug and alcohol testing part of the evaluation procedure, too, to see if you qualified. And of course, a lot of people would would file that.

 

Jenny [00:54:38] So the recommendations would be that went back to the two judges was that, you know, a judge would say, can you please look at person A I think they'd be a good candidate for electronic monitoring. The report would come back from on their suitability and say, we don't think so. And so judges stopped using it because they were no longer kind of on the same page in its purpose and how it should be used. So it just sort of disappeared from the landscape towards the end of 2007, and now it's hardly ever used, which is really unfortunate.

 

Jen [00:55:14] Yeah, that's so sad. All right, new campaign. We need to bring it back. This is going to be how we'll try to rally the troops here to bring it back to New South Wales.

 

Jen [00:55:26] Okay, last question. What's the research frontier? What are the next big questions in this area or related to this this topic that you and others will be thinking about going forward?

 

Jenny [00:55:36] Well, Jen, I mentioned at the beginning, I've worked on substance use quite a lot and I intends the next big question. I'm no good at predicting the future. So I take my interest, take me so I'm a bit of a wanderer and I've always been interested in substance use and criminal justice. And so I'm working a little bit more at this sort of intersection of those. Right now, I'm using beautiful administrative data I'm working with on In Bright Future Senate, the Norwegian Institute of Public Health and I'm looking at people who have entered substance use treatment in Norway, and I'm looking at in particular the impact of having to wait for access to treatment and how waiting time impacts a bunch of outcomes, including their criminality.

 

Jenny [00:56:25] And and also, you know, who they mix with if there are inpatient treatment, who they mix with, how that impacts their outcomes. So yeah, I guess I'm moving more towards the bringing together these two strands of literature I really care about. And I think that it's looking at substance use and it's a mental health problem a lot of people who are into substance use treatment have co-morbid conditions with other mental health conditions. People who are involved with criminal justice, mental health and substance use is a very much overrepresented in this group. And so I think that for me, if you're really looking at reducing reoffending programs that seek to address those conditions or circumstances that might precipitate their re-offending is also a really good place to look. So I think that, you know, looking at punishments and how people are punished, where they're punished, you know, the new prisons and are less harsh.

 

Jenny [00:57:27] The Norwegian prisons really do, you know, have programs which see people transition into release more gradually and that they have really good results from having those sort of programs also open programs. It's like lots of different ways of punishing people which are less punitive and a more effective at reducing reoffending, but also looking at the circumstances and situations that precipitate reoffending where interventions can be very helpful. And we know that the people with mental health problems and substance use a really high overrepresented in prison populations and offender populations more broadly and so looking at this intersection is where I'm currently working on.

 

Jen [00:58:09] I will look forward to seeing the results from that project. Very interesting.

 

Jen [00:58:14] My guest today has been Jenny Williams from the University of Melbourne. Jenny, thank you so much for talking with me.

 

Jenny [00:58:19] Thank you, Jen. Thanks so much. I really enjoy being able to share this. I hope people find it interesting and I hope there's policymakers out there who want to pick it up and run with it.

 

Jen [00:58:33] You can find links to all the research we discussed today on our website probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show and thanks also to our patrons, subscribers and other contributors. Probable causation is produced by Doleac Initiatives, a 501(c)3 nonprofit, so all contributions are tax deductible. If you enjoy the podcast, please consider supporting us via Patreon or with a one time donation on our website. Please also consider leaving us a rating and review on Apple Podcasts. This helps others find the show, which we very much appreciate. Our sound engineer is Jon Keur with production assistance from Nefertari Elshiekh. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.