Probable Causation

View Original

Episode 58: Jennifer Doleac

Jennifer Doleac

Jennifer Doleac is an Associate Professor of Economics and the Director of the Justice Tech Lab at Texas A&M University.

Date: September 14, 2021

Bonus segment on Professor Doleac’s career path and life as a researcher.

A transcript of this episode is available here.


See this content in the original post

Episode Details:

In this episode, we discuss Prof. Doleac's work on DNA databases:

“The Effects of DNA Databases on the Deterrence and Detection of Offenders” by Anne Sofie Tegner Anker, Jennifer L. Doleac, and Rasmus Landersø.


OTHER RESEARCH WE DISCUSS IN THIS EPISODE:


TRANSCRIPT OF THIS EPISODE:

Amanda [00:00:08] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your guest host, Amanda Agan of Rutgers University, where I am an Assistant Professor of Economics. You may remember me from such podcast episodes as episode eight of Probable Causation or Episode 51 of Probable Causation, but today the tables are turned and I get to ask the tough questions. My guest this week is someone you just may be familiar with - Jennifer Doleac. Jen is an Associate Professor of Economics and Director of the Justice Tech Lab at Texas A&M University. Jen, welcome to - well, your own show. 

 

Jennifer [00:00:47] Hi, Amanda. Thanks for having me. 

 

Amanda [00:00:49] Great. So today we're going to be talking about your research on how DNA databases affect recidivism. But before we get into that, could you tell us about your research expertize and how you became interested in this particular topic? 

 

Jennifer [00:01:03] Sure. So I am an economist by training and I now study crime, and I'm very interested in recidivism, particularly so how to stop the endless cycle of people through jail in prison. But actually first got interested in DNA databases when I was in graduate school and I was not committed to studying crime yet. So my job market paper was actually on DNA databases in the United States. And so at that time, I was just sort of looking around for natural experiments as econ graduate students do. And I found a great New York Times story about all of this crazy variation in state to state about how much DNA database laws vary from one place to another in terms of which types of offenders are required to provide DNA to the criminal database in that state. And I thought, oh, that's a great natural experiment. I wonder what we know about the effects of DNA databases, and it turns out we knew basically nothing. So that became my job market paper, and I'd been interested in the topic ever since. 

 

Amanda [00:02:06] Yeah. Great story about finding a topic for a job market paper also. So this paper that we're talking about today is titled "The Effects of DNA Databases on the Deterrence and Detection of Offenders." It is coauthored with Anne Sofie Anker and Rasmus Landerso and it is forthcoming at the American Economic Journal: Applied Economics. Congratulations, by the way. 

 

Jennifer [00:02:27] Thank you. 

 

Amanda [00:02:28] What are DNA databases exactly, and how are they used in the criminal justice context? 

 

Jennifer [00:02:35] Yes, so many people are probably now familiar with DNA databases like 23 and me and these various, like a family genealogy genetic databases where you can send in your DNA sample to some company and they, you know, you can find long lost family members or something. So that is not what I'm talking about. So I'm focused on government DNA databases and particularly the offender databases that all U.S. states and many countries keep. And so these DNA databases basically consist of two parts. So one is the offender DNA database and the other is the is a database of DNA samples from crime scenes. And so basically, if you are a criminal offender who is required by your state or national law to provide a DNA sample at the time of your charge or conviction, you provide a simple saliva sample. And at this point, it's all totally automated. So they'll put that saliva sample into a machine. And then a little while later, it'll spit out basically an identifying string of numbers. You can think of that as being akin to a Social Security number in the sense that none of the individual numbers actually mean anything, but it is unique to you. And so then that string of numbers is uploaded to the state DNA database. And then at this point in the United States, all these state DNA databases are connected into a network called CODIS by the FBI. And periodically, those offender DNA strings are compared with DNA strings from DNA collected to crime scenes. And so in that way, they can match offenders with crimes they've committed or where they should be a suspect, perhaps for the crime, in cases where they would not otherwise have been a suspect. So, you know, if you were already going to be a suspect in a crime, you don't need the DNA database to lead the police to you. But the idea here is that this can help connect offenders to their crimes in situations where they might not otherwise have been on the police radar. 

 

Amanda [00:04:31] Got it! And how widespread are these government DNA databases currently? 

 

Jennifer [00:04:36] So they're pretty widely used. In the United States, every state has a DNA database, and every state requires anyone convicted of a felony to provide a DNA sample to the state database. And all of those state policies are legislated at the state level. There isn't like a national database law, aside from just clarifying what the standards should be for being connected to CODIS. In addition, many states now require people who are charged with felonies or convicted of more minor offenses to provide a DNA sample to the state database. And then these databases are also widely used around the world. I'd say most countries at this point have one. This has been pushed quite heavily by our own FBI and the U.S. government supports the implementation of these databases elsewhere. So a lot of other countries actually use the same software that the FBI uses in its own DNA database. And every once in a while, there's a conversation about linking these databases across countries and being able to, you know, because offenders are increasingly mobile across countries. But that becomes a very complicated, very quickly. So I don't think, yeah, that doesn't happen in any official way at this point. But yeah, you should think of these databases as being very common at this point and widely used by law enforcement. 

 

Amanda [00:05:54] And why might these DNA databases affect crime and criminal behavior? Like, what are the mechanisms we should have in mind for how their adoption might affect crime and criminal outcomes? 

 

Jennifer [00:06:05] Yeah. So there are two main ways. One is if you have someone who is actively committing crime and who's added to the DNA database, then we could potentially catch them more quickly and take them off the streets more quickly. So if you have sort of a chronic offender who is a real public safety threat, then they add you - add them to the DNA database and we connect them with their next crime more quickly than we might have otherwise. And we can put that person in prison and prevent crime in that way. So there's an incapacitation effect that could result. The other way that these DNA databases could affect criminal behavior, though, is, you know, if offenders know that that's likely to happen, then once they go into the DNA database, they might be less likely to commit crime in the future. So we've essentially increased the probability of getting caught for future crime, and that could have a deterrent effect on future criminal behavior. And so we might see less crime because people are now deterred or we might see or if they're not deterred, then we can catch them more quickly and and lock them up and reduce crime that way. 

 

Amanda [00:07:07] So before this paper, what had we known about how DNA databases impact crime, I guess, from your own previous research and maybe from others? 

 

Jennifer [00:07:16] Yeah, so, so not a whole lot. It really was my own job market paper. So that was the main paper that was preceding this one. And in that paper, I had data from Department of Corrections in seven states. So it was able to look at the effects of DNA databases on recidivism in seven states in the U.S. and the policy change there that I was interested in was what happened when states expanded their database to include additional sets of felony convicts. So, you know, adding people convicted of robbery or burglary or something like that. And those policy changes were basically in the mid 90s or mid 2000s. And then I also had a part of the paper where I use these database expansions across the states to look at what happened to crime rates in the states. And so basically in that paper, I found that adding additional people to the DNA database reduced subsequent recidivism. So it did seem to have a deterrent effect both for violent felony conduct and for property felony convicts. And it also reduced expanding these DNA databases over time reduced crime rates. Now that all comes with a huge caveat that the data were super messy. So I'm very proud of that paper. But it was, I think, especially the recidivism piece. It's very difficult to get data in the U.S. that includes actually whether someone is added to the DNA database and when. So I had to impute all that stuff based on the person's sentence and the state law. And just for a variety of reasons, the data were kind of a mess and it was super noisy. So it found significant effects, I believed them. But it was also like, you know, I'll definitely want to see this replicated. And so this paper actually one of the main things that came out of it, this new paper we're talking about today is that it really convinced me that the facts that I found in the U.S. were probably really real because it's just so much cleaner in this other context that we're going to be talking out. 

 

Amanda [00:09:16] Yeah. So it sounds like in some sense, this is a difficult question to answer. There are many things that might make it a little difficult to try to get at the impact of DNA databases on crime and recidivism. You just mentioned one of them about data challenges. Are there any other challenges that one faces in trying to answer this question? 

 

Jennifer [00:09:34] Yeah. So the data one is definitely a major barrier, especially if we want to study this in the U.S. But the other piece is really the identification challenge, right. So you don't want to just compare recidivism or criminal trajectories of people who are added to the DNA database to people who are not. Because in general, just by definition and by law, the people who are added to the DNA database are more serious offenders or have a different criminal history than people who are not added. That different criminal history could drive any subsequent differences that we see in criminal behavior. And so you really need some sort of experiment. And in this case, it's very unlikely to be an actual randomized controlled trial. We're not going to actually randomize who we put in the DNA database, just like we wouldn't actually randomize who we take fingerprints from or lock up or whatever else we might do in a criminal justice system. So we need a natural experiment that does something similar that kind of sought similar people into a group that is added to the DNA database and a group that is not. And those natural experiments are always a little tough to come by. But that's the main challenge in this case. 

 

Amanda [00:10:43] Great. So to overcome some of these challenges, you're going to leave the United States context and head over to Denmark and use a big expansion of DNA databases in Denmark as a natural experiment. Can you tell us about this policy reform? 

 

Jennifer [00:10:55] Yes, the main draw of Denmark is the data. But they also, you know, the other piece we needed was some sort of policy change that provided this natural experiment, and luckily they had one. So back in 2000, Denmark created a DNA database. At that time, it included very few people, so they basically only included a very, very serious offenders and only if DNA was relevant to the case. So that additional restriction meant that even a lot of people convicted of rape or murder would not be included in their DNA database because, you know, maybe they caught the person red handed or they confessed or for whatever other reason, DNA was just not part of the investigation. So then fast forward to 2005, and they expanded this database to include anyone charged with a crime carrying a potential sentence of one and a half years or more. So you can think of this as roughly equivalent to anyone charged with a felony in the U.S.. So that policy change went into effect on May 24th, 2005. And basically, it means that there's just a massive expansion in the number of people added to the DNA database and the probability of someone charged with any offense in Denmark being added to the database went from about four percent, so very few people added to the database before this policy change to 40 percent by the time it was fully phased in in October of 2005. 

 

Amanda [00:12:17] And how are you going to use this policy reform to measure the impact of DNA databases on behavior? 

 

Jennifer [00:12:23] Yeah. So the basic intuition here is we're going to compare people who were charged with very similar crimes just before the database expansion went into effect. So think you know early May 2005 with people who are charged with essentially the same crime or the exactly the same crime after the database expansion went into effect. The idea here is you have two very similar people and just by sort of luck of the draw. One happened to be charged before the database expansion and one happened to be charged after the database expansion and everything else we can control for sort of other trends going on over time. And just in general, these two people are going to have the same experience of the criminal justice system in the world. You know, they're living in about the same time they're just their charges happened at slightly different times. And so we're going to be comparing those two people and the person who's charged just a little earlier doesn't go on the DNA database and the person in charge just a little later does go in the DNA database. And so we can use that to follow those two people over time and see what the effect of being added to the DNA database was on subsequent criminal behavior. That's a little more complicated than this. As always in Denmark, it turns out that people take a real summer vacations and that includes both criminals, which is interesting, but also the police. 

 

Jennifer [00:13:35] So this policy change happened at the end of May 2005, and then everyone went on summer vacation. And so the implementation of this law was actually quite slow. And so there's a nice graph in the paper. We kind of show the share of people charged out of the database goes up, but it's sort of like goes up more slowly during the summer months and then by fall, it's really in effect. So have to do some stuff in the paper where we basically in most of our analysis, drop those summer months as are worried about selection effects about who actually makes this database. Then it turns out it doesn't really matter if we do that or not. So what we're actually doing is then comparing people charged before May of 2005 and people charged October or later in 2005. And so if you do a whole bunch of the paper to kind of convince ourselves that that's not a problem. But so what we're doing here is basically going to use the date of the charge and instrumental variable for whether DNA was added to the database. And for those who know the space you can think of, what we're doing is akin to a donut RD or a donut regression discontinuity. But the basic intuition here is were comparing similar people charged on either side of this database expansion and seeing what happens to them over time. 

 

Amanda [00:14:42] All right, you already alluded to it, but tell us about this amazing data from Denmark that you're able to use for this paper. 

 

Jennifer [00:14:49] Yeah. So Denmark is one of these countries that just, you know, collects all these data on everybody and links it all together. And then it's available to researchers. And so we have information on people's full criminal histories for everyone who's a Danish citizen. And then we're able to link that for some of the analysis in the paper to information about employment and education and family structure, where they're living and that sort of thing. And then also the other nice piece here that I didn't have in the United States is that the criminal history data included information on whether and when you were added to the DNA database. And so that sort of really nice piece that allows us to, you know, measure these effects much more precisely than I could in the U.S. And we also have just really rich information on the particular offenses that people are charged for later. And we're going to use that in various ways to try to tease apart mechanisms and so on. But yes, we have really rich data on basically everything we'd want to know about their criminal history. And then some other stuff. So we're going to focus on people charged between 2003 and 2004. We're going to basically two years pre and post the law change dropping those summer months of 2005. Our sample we're going to restrict to men who are age 18 to 30 at the time of the charge. We want to drop juveniles basically because you know that - just they're just different a bunch of ways. And you know, those 18 to 30 year olds are the ones who are most criminally active, so they're the ones who are most interesting. The other piece that is somewhat important is that we have to restrict our analysis to Danish citizens because they're the only ones that are really tracked in this database well. So that means we can drop any tourists and any people who are just coming through the country who are EU citizens. Turns out, tourists commit not an insignificant share of crime in Denmark, and so it's not trivial. But obviously most of the crime is committed by Danish citizens, so it's not too much of a cost. 

 

Amanda [00:16:36] All right, so you've got this great policy change. Great empirical strategy. Amazing data. Put it all together. What do you find? What was the causal effect of being added to the DNA database on recidivism? 

 

Jennifer [00:16:50] Yeah. So we find that for those people who were charged just after that expansion went into effect, so they were addded to the database and now are much more likely to get caught for their future crime right, so thid probability of getting caught has gone up, we find that the probability of a conviction for a new crime, for those people within a year after that initial charge falls dramatically by 42 percent. We also find the number of new - of convictions for new crimes also falls by 49 percent, so about the same size. And these effects, which are quite large, persist for at least three years. And you know, we might expect them to to get smaller over time, as everyone who has was not in the database yet continues to get arrested and charged may be added later. So we can't go too far out. But basically we see huge effects on recidivism that lasts at least three years. 

 

Amanda [00:17:39] Did you see different effects across different groups of offenders when you looked at that? 

 

Jennifer [00:17:44] Well, yes and no. So for the most part, basically everyone's affected by this, with one exception. So we find that the effects were entirely driven by younger offenders, even within this relatively young group of 18 to 30 year olds. The effects are really entirely driven by those aged 18 to 23. So basically the bottom half of that. But otherwise. Everyone is affected by this, so we see the strongest effect for people who were initially charged with a violent offense. But we also see large effects for property and weapons offenders. We see larger effects for first time offenders and for recidivist, but they were significant and large for both. So just kind of, however, we might want to cut it. We're seeing effects for every group, except that slightly older group within our within our sample. 

 

Jennifer [00:18:27] So it is, I think, interesting that this is affecting - I think a lot of people think of DNA is based as really being only relevant to people who are committing violent crime. You need a big pool of blood or something to - for that DNA sample to be for you to collect DNA at this crime scene. And basically, the technology's just evolved at this point that it's really relevant for any type of offense where your presence at the scene could be evidence that you might have been involved, right. So a burglary, you know people are touching things constantly when they're in a house, and so you can just collect very small DNA samples and be able to analyze that and potentially match it to an offender. And so you can think of this as being - you can think of DNA as being relevant to a whole lot of property crimes in addition to violent crimes. And so that helps explain why we might see a broad range of offenders that are affected by this. In addition to possibly just the story that I think I have in mind about this, that, you know, you're added to the database and essentially at that point you decide, okay, I need to really, you know, straighten up my act. If I keep doing the stuff I'm doing and hanging out with the same people I'm hanging out with, I'm going to get caught really quickly and wind up back in prison. So you just sort of clean up your act more broadly rather than only avoiding crimes where DNA is likely to be collected. But yeah, so it does seem to be really effective for a really broad range of offenders. 

 

Amanda [00:19:45] Yeah. So you just touched on this a little bit, but what types of new crimes are prevented by this reform, all of them or a certain subset? 

 

Jennifer [00:19:53] Yeah, it's really all of them. So we see big reductions in subsequent - so we could think about like what types of crimes are people not committing anymore? Like what types of crimes are prevented? And it really is kind of across the board. So violent offenses, property offenses, weapons offenses all fall substantially and significantly after someone's added to a DNA database. 

 

Amanda [00:20:13] Yeah. And you talked before about all this sort of data that you had from Denmark. And in that case, you're able to also consider non crime outcomes, things like employment, education, family structure. What types of effects of this DNA database expansion did you find on those outcomes? 

 

Jennifer [00:20:29] Yeah. So we were really interested in to what extent, you know, being able to put someone on to a better track, maybe or push someone onto a better track. And can you know, if we think of basically the effect of adding someone to the DNA databases as making a criminal career less appealing to them that could have effects on a whole bunch of areas of their life, right. And if this is really changing the amount of crime that someone's committing, then we might expect to see that show up in other areas of their life as well. So we might see them be more likely to be working or enrolled in school or settling down with a family or something like that. And so because we have this rich Danish data, we're able to link to all those things and actually test this hypothesis and see what other areas of their lives seem to be affected. And so we do see effects on unemployment and education. Story is a little bit different depending on age. So younger offenders seem to shift from being unemployed to being enrolled in some sort of education or training program. And older offenders, those that were 23 to 30, relatively older by criminal standards, they shift from unemployment to employment. So overall unemployment goes down. But what people wind up doing instead is different depending on their age. And we think, you know, especially for that younger group that's investing in education or training, it really suggests that they're now investing in a different career path. 

 

Jennifer [00:21:52] And then we also look at what happens to family structure, the likelihood of being married, of living with their child, that sort of thing find now much more suggestive evidence there. It's not quite as solid, but we do find some suggested evidence that especially for first time offenders who are out of the DNA database, they're much more likely to be married if they're added to the database than if they are not all the effects on family size, as their family structure tend to be large in magnitude but not statistically significant, suggesting that there's just a lot of variation here that's really tough to explain. But overall, we do see, at least, you know, evidence in line with the idea that these people are just now on a very different track once they are in the database relative not being in the database. Of course, we can't say anything solid about kind of, you know, what's causing what. It could be that being added to the database changes the way that people pursue a whole bunch of opportunities in their lives. And it has sort of a direct effect on, well, I'm going to go to school now instead of hanging out with these troublemakers that are always getting me to do dumb stuff that gets me arrested. And so that that could be a much more direct effect. Alternatively, it could be that because you don't commit crime anymore, you're less likely to be arrested. And thrown into jail or prison, and for that reason, you're able to stay in school or hold on a job. And so, you know, the direction of causality or how these things are causally linked is a little less clear about what we can say is that adding to people's the DNA database does have either a direct or indirect effect on these other outcomes as well. 

 

Amanda [00:23:23] Great. Like any good economics paper, you do a lot of robustness checks to convince yourself that the impacts that you're estimating are truly the causal effects of the DNA database. What tests in the paper do you find the most useful? 

 

Jennifer [00:23:38] I'm so glad you asked this, Amanda. I do have some favorite robustness checks in this paper. So as I mentioned earlier, one of the main challenges here was that summer, the summer months when everyone was on vacation and having to drop those summer months right after the policy change. And you know, we'd really want to be doing here is it's a really nice regression discontinuity where we compare people just before and after May 24th. And because we have to drop the summer months, it's just, you know, we do the best we can to convince ourselves that you know we're controlling for everything else that might be changing during that summer, but always in the back of our minds, we wondered it's just some weird seasonal effect that our design is picking up mechanically. And so the robustness check that convinced me that the results were real was we were able to do a placebo test where we basically ran the exact same analysis, but in other years. So imagine, you know, we pretend that the policy change happened in May 24th of 2004 or 2003 or 2007 instead and look for the same change around the, you know, drop the summer months, run the exact same analysis and we find just zero effects in all those other year. It's only in 2005 where there was a real policy change and there was a real difference in the likelihood that someone was added to the database where we see this dramatic drop in recidivism for people who were added to the database relative to those who weren't. So that placebo test, I think, is really nice and compelling and convinced me. 

 

Jennifer [00:25:05] The other thing we can do is a difference in difference analysis, where instead of the underlying assumptions for our instrumental variable strategy are that the date that you're charged, whether it's before and after this policy change, that date is affecting your likelihood of DNA registration. But we're assuming there is that subsequent effect on recidivism is happening only through that effect on DNA registration, not some other, not something else that's going on. And we might worry about that for, you know, a variety of reasons, especially this seasonal issue. And so instead, we can do a difference in difference where we basically break people up, bring charges up into categories that are less affected by the policy change, where a relatively small number of those offenders are going to be added to the database and charges that are more affected by the policy change where the vast majority of offenders charged with that crime are going to be added to the database. And so it's essentially comparing groups with different intensities of treatment from this policy change. And so we would expect there to be a bigger effect on the group that is more affected by the policy change. And so we can compare these two groups and the recidivism rates across these two groups over time before the policy change. And they're super parallel, which is really nice. And then once we went, the policy change happens. We see a gap in recidivism rates starts to emerge, which is very much in line with our overall results from the the other strategy showing that people who are added to the database do begin recidivating less. Because that's kind of nice because we can basically show the same results, but with a different set of underlying assumptions for the causal effect. 

 

Jennifer [00:26:38] And then the last thing we do is say, well, you know, if this effect is real and recidivism is really dropping by 40 percent for this, this huge group, then we should see this effect in crime rates right, at least, if not if there isn't like a flood of new offenders entering entering this market to take their place. This is a big enough effect that we should probably see the effect in actual numbers of reported crimes. And so we can look at that, we can look at, you know, what happens reporting crime rates before and after this reform. Now this is not, you know, an analysis that we would want to lead with in the paper. There's a bunch of other stuff going on in the country. And, you know, we don't have a good control group here, but we do see a big drop in crime rates after the policy change. So it's very much consistent with the idea that this is a real effect on criminal behavior and not some just weird artifact of our empirical strategy. Or even, you know, maybe people are just getting better at avoiding criminal detection or something like this is just seems to be a real drop in actual crimes being committed. 

 

Amanda [00:27:40] So one challenge we always face when studying crime and criminal behavior is that it's impossible to get data on true underlying criminal behavior, at least usually. We need to rely on administrative databases and when somebody gets caught, when somebody gets arrested or charged. And so as you note in the paper, when challenged then to studying effects of tools like this because they work exactly by increasing the likelihood of getting caught, even if underlying criminal behavior doesn't actually change, an offender could be more likely to show up in the data in the future. So for this reason, the estimated effects on recidivism could be biased upward because data quality has improved or more likely to match you to your crime. This means that the true effect of DNA databases on recidivism could be bigger than the numbers you just discussed. To look at this, you take advantage again of this rich Danish data you have to dig into this a bit more. What do you guys do and what do you find? 

 

Jennifer [00:28:36] Yes. So this is a place where we take advantage of the really wonderful Danish data to be able to do two things. One is that we are able to take advantage of the institutional fact that if you collect DNA from a crime scene and have it analyzed and then uploaded the DNA database where it could potentially match your offender and then have that information be sent back to law enforcement, all that's not going to happen in an hour. Right. So despite what we've seen on crime shows on TV, this stuff takes a while. And in fact, in Denmark, it takes about four weeks for that to happen. That means that offenses where someone was charged with that crime very quickly after the offense happened. We know for sure that there's no way the DNA database led police to that suspect. It's just impossible in that case, so they must have solved the crime in other ways. And of course, you know, police solve crimes in other ways all the time. They're witnesses that saw somebody or it's on, you know, surveillance video or they're caught in the act or it's just an obvious suspect, right. If a wife gets murdered, then the husband will be the obvious suspect. You don't need the DNA database to find the person. And so it turns out that the vast majority of crimes, if there's someone caught, they're charged pretty quickly after the offense. It doesn't get to that three or four week mark where the DNA database might possibly have assisted in identifying a suspect. So it turns out that in Denmark, we have really rich information on the timing of the offense and the timing of the charge. And so we're able to say like, okay, well, let's just look at the likelihood of for these people that are charged, you know, have this initial offense and added to the database or not, the pair those two groups over time and see if they're charged and convicted of future offenses. What we can do is say, are they convicted of a new offense where the charge happened really quickly or where it took longer, where it took three weeks or more to be filed? And so in those in the offenses where the charge happened really quickly, the only way that criminal behavior could be affected there is through a deterrent effect, right. Because there's no way that there's the detection effect of DNA databases that's layered on top. 

 

Jennifer [00:30:45] So we're not worried in those types of cases that we're simply seeing your criminal behavior more often because the database has led us to you. And so we're able to say okay for the types of charges where we see a charge happening really quickly after the offense, only a deterrent effect can be operative there. But in offenses where a charge came more slowly, it took, you know, three or four weeks, we cut it at three weeks to be a little more conservative. It usually takes four weeks or more. Then it could be a combination of a deterrent effect that pushes the likelihood of seeing you in our subsequent data down and a detection effect that pushes it back up and that's that upward bias you were talking about. And so it's a combination of those two effects. And so basically, we use this kind of institutional fact about how long DNA analysis takes along with it's really rich Danish data on the exact timing of the offense and the charges that people are caught for later to be able to separate these two effects - these deterrent and detection effects. 

 

Jennifer [00:31:40] And so what we find is, in this case, most of the effect that the DNA database is in fact coming from deterrent. That's the overwhelming effect here is is coming on these charges that happen really quickly after the offense where there's no way the database could have led a police to you. But we do find that there's a significant detection effect. So being added to a DNA database increases the probability of detection for a new crime by four to five percent. So, you know, that's not nothing and certainly means that if we weren't thinking about this, then our results would be biased and we could potentially miss important deterrent effects if we were only looking at sort of the aggregate of the two. 

 

Jennifer [00:32:16] And the other nice thing we're able to do once we separated these detterent and detection effects is measure or estimate the elasticity of crimes committed with respect to that detection probability. So we find that for every one percent change in the probability of getting caught, we find a 2.7 percent reduction in the number of new crimes committed. So that's a big effect, not totally out of range, with estimates coming from studies of what the effect of hiring additional police officers say on the likelihood of people committing more crime, but certainly is in line very much in line with an idea that gotten a lot of more traction in research and policy circles recently that increasing the probability of getting caught has a much bigger deterrent effect on crime then increasing the punishment, say, putting people in prison for a really long time. And so this estimate of this elasticity is very much in line with that idea that this tool is working by increasing the probability that you'll get caught if you commit more offenses, and that just has a pretty dramatic effect on the likelihood that you will re-offend. 

 

Amanda [00:33:22] All right, Jen, the million dollar question, what are the policy implications of this paper and your other work in this area? What should policymakers take away from these results? 

 

Jennifer [00:33:33] I think one big policy implications these databases are really work. You know, I think going into this whole line of research, I was pretty agnostic about whether this type of tool would have the big effects that a lot of its proponents promised. And certainly in a lot of spaces, I mean, I study a lot of technologies in the public safety space, and often they are really oversold, I think. And like, that's, you know, they always have a ton of potential. But then as they're implemented, it's kind of like, well, I don't know if we're really getting as much as we hoped for for that. But in this case, we're really getting a really big bang for our buck. And the buck is, you know, relatively small, like once you invested in the database infrastructure and the DNA crime labs and all that stuff. And we had at this point, all that stuff exists. And the question is, should we be expanding the DNA database to add more offenders? And the answer to that seems just from a purely financial perspective, like a clear yes, the cost of that is really cheap. The gains we get from it in terms of reduction of crime are really big. So it's just incredibly cost effective just from a pure like how do we reduce crime perspective. Now, the big concern that a lot of people listening might have is, well, what about privacy, right. Is you're collecting DNA samples from people and the government is collecting DNA samples and analyzing it and adding you to a big DNA database. And that just freaks people out. And I think that is honestly like the biggest potential cost that we need to think about as a society. 

 

Jennifer [00:35:02] Now, I think there are a couple of things that, to me, make DNA databases feel less intrusive than a lot of the alternatives. One is yes, your DNA is being collected, but the government's not analyzing it for a whole bunch of like health information or really sensitive information. It's not the same as when you send a DNA sample to one of these genealogy websites or 23 and me or whatever, and they're giving you a whole bunch of information about your health and what you're risk for and all that stuff. They're doing a whole lot more with your DNA than what the government's doing. The government is really just putting together this string of numbers. You can think of it as being just a more accurate fingerprint. And so to the extent that this is just a more accurate fingerprint than we already fingerprint everybody who comes to the criminal justice system, you know, it seems better to have a more accurate fingerprint that also is very salient to people and seems to reduce their future criminal behavior so much. 

 

Jennifer [00:35:54] And the other thing is that, you know, the alternatives are often much more invasive. So if you think about other ways that we might be able to increase the probability of getting caught, they all involve some level of sort of keeping tabs on people. Right. So you got cameras everywhere or you can have police following you around and like, you know, frisking you all the time or you can actually put someone into prison, right. And that's a really good way to make sure that they're not causing any trouble. All of those things have really large privacy costs. And I think of those the privacy costs for those alternatives as being much higher than for having a DNA sample analyzed and have a string of numbers uploaded to a database. Now, reasonable people can disagree about this. And I think that is that's the conversation we need to have as a community, as you know, whatever you perceive, the privacy cost to be is it worth this dramatic drop in criminal behavior? And some people will say yes, and some people will say no. 

 

Jennifer [00:36:50] The one other thing I will add to that is there already are communities that are surveilled in an extreme way all the time where cops are following them around and they, you know, there are big invasions of privacy all the time. And it might not be the worst thing in the world if we extend some of that concern to a broader community so that we all have to focus a little bit more on making sure that we have the protections we want in the criminal justice system. And you aren't going to be wrongfully convicted if you do happen to leave your DNA at a crime scene just because you were, you know, you walked into a store that morning and bought a soda before it was robbed later in the afternoon or something like that. So I think to the extent that this just makes it more salient to people that, you know, we need to have the safeguards in place in our criminal justice system to make sure that people aren't wrongfully convicted all the time. I think that might not be the worst thing in the world.

 

Amanda [00:37:46] So you talked about your job market paper and this paper with this policy reform in Denmark. Have any other papers related to this topic come out since you guys first started working on this study? 

 

Jennifer [00:37:56] So not on DNA databases. I have a work in progress with Andie Kelly and Venetia Shaw, where we're looking at the effects of government grants to help clear rape kit backlogs on what happens to samples that go into DNA databases. But there really isn't much else on DNA databases specifically. Now, more broadly, there have been other papers looking at the effects of surveillance cameras. They generally find reductions in criminal behavior. There's some other work like that that's sort of in this space of what's the effect of technology on public safety. More generally, I've become much more interested in this question of how to break the cycle of re-incarceration that we see in the U.S. So about, you know, about half of people who are released from prison are going to be re-incarcerated in three years. And so, you know, we just need to do a much better job of finding ways to reduce criminal behavior. It seems like DNA databases are one really effective way to do that. A lot of other things that we try don't work. And so I've written a review article now kind of trying to just pull together what we know about what is effective in that way. And there are certainly lots of great ongoing work in that space that speaks to just how to reduce recidivism more broadly. And so that is probably the most relevant stuff. But on this question of how does this or other technology affect recidivism, there's still a lot we don't know about a lot of high tech tools that are out there, and I think it's a great area to be working in. 

 

Amanda [00:39:30] Yeah. So what is that research frontier like within that? What are the next really big questions that you think somebody could tackle? 

 

Jennifer [00:39:36] Yeah. So one is just to find one of the myriad high tech tools out there and see if you could figure out a way to to measure what the effects are. And the other big question that comes out of this paper is, you know, one thing that we couldn't do much about is that, you know, we're able to measure in that last piece that I was talking about, we're able to look at what's the effect of an increase in the probability of detection for crime on the likelihood that you will re-offend. So thinking about that elasticity of crime with respect to the probability of getting caught. What we're doing there is using the actual change in the probability of detection, like when we run the numbers, like what is the increase in the likelihood that you're more likely to get caught when when you're added to the database? 

 

Jennifer [00:40:22] Now what we're not able to see is what people perceive as the change in the likelihood that they're going to get caught. And my hunch is that people overestimate the likelihood that being added to the DNA database means they're going to be caught for a future crime. And this is because, you know, most offenders are probably not learning what they know about DNA databases from like scientific journals, they're watching CSI or whatever the hot crime show is these days. And so they're imagining on added to the DNA database. Now, the next time I go rob a 7-Eleven, my photo is going to appear in the sky like every police officer will like, know who I am and be able to find me in 10 minutes. And that's not the way DNA databases work. So to the extent that people are overestimating the likelihood that they're getting caught, that they're going to get caught if they re-offend, then then that would affect, you know, this elasticity that we're estimating it will be biased upward a little bit. 

 

Jennifer [00:41:18] So in terms of research frontier, I think having some way to measure or estimate offenders of perception of the likelihood of getting caught, I think will be important for this and a variety of other contexts, especially as we move further in this direction, which I think we will of really focusing on increasing the probability of getting caught rather than increasing or relying so much on incarceration or punishment. And in general, I think that will be a good move. But what we really want to know is what offenders think that likelihood is because that's what we really want to target to the extent that it's highly correlated with the actual probability, that's great. But I could imagine places where just for a variety of reasons it's not going to be perfectly correlated. And so figuring that out a little bit better will be helpful. And then there's just, you know, there's so much we don't know about how to reduce recidivism more broadly. If for whatever reason, you really don't like the idea of expanding DNA databases, then we need to do something else. And there are so many well-meaning programs and policies out there in the space that do not work. And so we just need a whole lot more research and experimentation and evaluation on other interventions that could help people reintegrate into their community successfully after they get out of prison or after they're convicted of a crime. And we have a lot more work to do on that front. 

 

Amanda [00:42:37] Well, this has been a really interesting conversation. My guest today has been Jennifer Doleac from Texas A&M University. Jen, thanks so much for talking with us. 

 

Jennifer [00:42:45] Thanks, Amanda. 

 

Jennifer [00:42:52] You can find links to all the research we discuss today on our website probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show, and thanks also to our Patreon subscribers. This show is listener supported. So if you enjoy the podcast, then please consider contributing via Patreon. You can find a link on our website. Our sound engineer is Jon Keur with production assistance from Haley Grieshaber. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening, and I'll talk to you in two weeks.