Episode 91: Allison Harris
Allison harris
Allison Harris is an Assistant Professor of Political Science at Yale University.
Date: April 11, 2023
A transcript of this episode is available here.
Episode Details:
In this episode, we discuss our work on increasing the civic engagement of people with criminal records:
"Registering Returning Citizens to Vote” by Jennifer Doleac, Laurel Eckhouse, Eric Foster-Moore, Allison Harris, Hannah Walker, and Ariel White.
OTHER RESEARCH WE DISCUSS IN THIS EPISODE:
“Can Incarcerated Felons be (Re)integrated into the Political System? Results from a Field Experiment” by Alan S. Gerber, Gregory A. Huber, Marc Meredith, Daniel R. Bigger, and David J. Hendry.
“The Politics of the Restoration of Ex-felon Voting Rights: The Case of Iowa” by Marc Meredith and Michael Morse.
“Using Causal Forests to Predict Treatment Heterogeneity: An Application to Summer Jobs” by Jonathan David and Sara B. Heller.
"Estimation and Inference of Heterogeneous Treatment Effects using Random Forests" by Stefan Wager and Susan Athey.
“Civic Responses to Police Violence” by Desmond And and John Tebes. [Working Paper].
“Mobilized by Injustice: Criminal Justice Contact, Political Participation, and Race” by Hannah L. Walker.
Bonus Episode 10 of Probable Causation: Hannah Walker.
TRANSCRIPT OF THIS EPISODE:
Jennifer [00:00:08] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an economics professor and the director of the Justice Tech Lab. My guest this week is Allison Harris. Allison is an assistant professor of political science at Yale University. Allison, welcome to the show.
Allison [00:00:26] Thanks so much Jen, it's great to be here.
Jennifer [00:00:28] Today, we're going to talk about our research with several coauthors on increasing the civic engagement of people with criminal records, but before we get into that, could you tell us about your research expertise and how you became interested in this topic?
Allison [00:00:42] Sure. Sure. So in the area of American politics, mostly I'm interested in law and courts. So all of my research is somehow related to what happened in the court system, in the legal system, more broadly and typically, when I'm conducting this research, I'm interested in issues related to the criminal legal system. So about how criminal trial judges sentence and make decisions about traffic stops. I've done work on disparities in traffic stops and also just how we think about this system, you know, really broadly in terms of all the players involved. So this project here is a little bit further afield for me because I'm usually thinking about what happens in political institutions, but obviously directly related to what individuals who interact with the system, who have contact with the system are going through.
Jennifer [00:01:32] So our paper is titled "Registering Returning Citizens to Vote." It's coauthored with a bunch of amazing people Laurel Eckhouse, Eric Foster-Moore, Hannah Walker and Ariel White. There are a bunch of us, which is why we're not all on this in this interview at the same time. So felon disenfranchisement has become a big political issue with reformers trying to change state laws that ban people with criminal records from voting. So tell us a bit about the legal landscape here. What happens to your right to vote when you're convicted of a crime?
Allison [00:02:03] So as with most things in the United States, there's a lot of variation here. And so each state has its own laws with respect to what happens when somebody is convicted of a crime, when somebody is convicted of a felony you know what happens to their ability to vote. And so, you know, in most states, after serving some or all parts of a sentence, whether that's just a part that includes incarceration, whether this goes through any sort of post-incarceration parole, for example, at some point after individuals complete all or part of their sentences, they usually regain the right to vote if it was lost at all in the process. There are some states that are super liberal and then, you know, sort of don't take away that right, but in most states you're going to have individuals in the system are going to have some loss of right to vote for some amount of time.
Jennifer [00:02:53] Yeah and I think this is usually when so it's while you're incarcerated.
Allison [00:02:57] Yes.
Jennifer [00:02:57] While you're under supervision, like on parole or probation.
Allison [00:02:59] Yes.
Jennifer [00:03:00] And then, as you said, most places, you're right, is automatically restored as soon as that sentence ends, but then, of course, you have to, like, know that and reregister. Right.
Allison [00:03:12] Exactly. Which is a big piece of the puzzle I think we're trying to solve.
Jennifer [00:03:15] Yeah.
Allison [00:03:16] Knowing that.
Jennifer [00:03:17] Yeah. So for those with past convictions who are eligible to vote, what do we know about their current voter participation rates?
Allison [00:03:23] They're pretty low, you know, compared to most other populations. People who have lost the right to vote, people and even people who have sort of been in the system and maybe haven't even lost the right to vote, typically have lower participation rates than individuals who have not had a lot of contact with the system. And especially then, you know, when we're thinking about individuals who've been convicted of a felony and have had to serve some sort of sentence.
Jennifer [00:03:48] So obviously than simply making people legally eligible, as I was just saying, automatically convert them into voters. So what are the reasons that those participation rates might be so low?
Allison [00:04:01] You know, I think a main reason what you just alluded to is that people aren't aware of their restored right to vote. So I think people become very aware that they've lost the right once they're convicted. I think most people even assume that if they are convicted of a felony, they will no longer be able to vote, but I don't think there is a lot of widespread information about when and how people regain that right after a conviction. So I think that's a huge lapse right there. It's possible that people would register and would vote if they knew they had the right to do so. Another barrier here is that people are often aware that if they attempt to vote when they're not eligible, in many places, this is another felony conviction. Right. And so if you've just been convicted of a felony, you probably don't want to risk another one, especially for something so seemingly minor as trying to vote compared to, you know, committing what we would think of as an actual crime.
Jennifer [00:04:59] Yeah, I think there's it's been interesting. We've had a lot of conversations within our research team about the various possible reasons here, right.
Allison [00:05:06] Yes.
Jennifer [00:05:07] I'm sure we'll get into this more as we kind of talk about exactly what we're doing in the intervention we designed, but sort of the other big hypothesis floating around out there is that the type of person who's had interaction with the criminal legal system just isn't interested in participating because their experience has been so negative. I think that's something political scientists think a lot more about than economists. So you yeah, you might want to see a little bit more than me.
Allison [00:05:33] We do there. You know, there is the line of thought here, the line of reasoning and research that looks into the possibility that right, like once individuals have contact, usually unwanted contact with this system, if this is their primary interaction with the state and for many people who get caught up in the system, this is in fact their primary contact with the state. Right. We're not talking about a group of people who are likely to have had meetings with their state representative, for example, or who are likely to have met one of their members of Congress. Right. This is the face of the state that they are most likely to interact with and have interacted with most often. So this is how you conceive of government if this is how you conceive of the state, you can imagine a world in which you wouldn't feel very much like participating in it voluntarily if all of your interactions with it have been pretty negative. So that's another reason.
Allison [00:06:26] And then sort of another reason on top of that is because it's kind of known that this group of people tends to be less likely to participate in these more formal ways politically campaigns don't necessarily reach out to them. Right. And so in political science, you know, one of our sort of longstanding findings about sort of voting and mobilization is that people vote because somebody asked them to. Right. So this is this is one way we think about it mobilization is really important to some people in terms of getting them out to vote. And if nobody's asking these people to vote because they assume that they won't vote this is kind of this reinforcing cycle potentially.
Jennifer [00:07:02] Yeah. And I will add one more hypothesis to this list, which is that we know that voting voting is often difficult in.
Allison [00:07:10] Yeah.
Jennifer [00:07:10] Many states in the U.S. right. So you have to have a valid ID in many states and you have to be able to take off time from work to go to the polls.
Allison [00:07:18] Right.
Jennifer [00:07:18] On Election Day. And this is a group that people who have a criminal record, especially a felony conviction, are likely more economically disadvantaged than your average U.S. citizen. And so it might just be that disadvantage that also makes people less likely to vote. So it might be that people with criminal records aren't voting at lower rates than sort of similarly economically disadvantaged people and we don't we just don't have great data on that.
Allison [00:07:45] Right exactly exactly. Right so it could also be the resource issue, right as you said. And we try to get at this and I'm sure we'll talk about it a little bit in our project, trying to sort of compare this population to a population of people who may be similar on other and least observable to us dimensions and see whether or not they are similarly easy or difficult to mobilize to register and vote.
Jennifer [00:08:11] Yes okay. And I guess I should also ask the question that I often get from my curmudgeonly economics colleagues when I present this paper around, why is it bad if this group engages at low rates, why should we want to increase their voter participation?
Allison [00:08:25] I mean, for me and I guess because I'm not a curmudgeonly economist.
Jennifer [00:08:29] Lucky you.
Allison [00:08:29] The answer to this is always democracy, right? So this is how democracy works. And it works when people are engaged everybody has the right and should have the right and be aware of their right. So if individuals are regaining their right to vote, then we feel best about our democracy if they're able to use it. I think for me, especially if we find that big gap is one of information, this is something that we can easily close. It's also an empirical question. We can try to figure out if the biggest gap here is sort of the main reason that people aren't registering because information this is a gap we can fill. We can identify that problem and we can actually come up with answers to it.
Jennifer [00:09:12] Yeah, I think there's also just a sense that and I don't know how much evidence there actually is for this, but I at least have a sense that our our democracy and our government works better when there is a representative group of the population that is participating. Right? So we want people.
Allison [00:09:27] Oh absolutely.
Jennifer [00:09:27] Who have that experience and have experience with all facets of society to have a voice and be engaging. And so if the group that has the most direct experience with our criminal justice system, especially when that criminal justice system is not working very well, if they're not at the table and talking with their representatives and having a voice in how to change it, that's going to lead to worse policy.
Allison [00:09:51] Exactly. And and, you know, we have to remember that at some stage at least, they're systematically they're removed from the process. Right. So these are prime stakeholders who are sort of forcibly removed from the process and then get sort of brought back into the process and so should therefore be able to behave as stakeholders. Right. And cast their votes to sort of create a system that they think better represents them. Absolutely.
Jennifer [00:10:18] Yeah. And then I'll add one more. Since I do get this question a lot, I've thought a lot about it. One more piece that especially comes from the efficiency side, which economists are looking for. You know, I think there is one motivating factor here in this study, and at least one of the main reasons I became involved in the study was that there is this idea a hypothesis out there that increasing civic engagement can facilitate the reintegration of people with criminal records back into society and perhaps reduce recidivism. And so if getting people to register to vote reduces recidivism, that is a really cheap way to break the incarceration cycle. And we're not going to get at that in this paper, but I still have hope of testing that hypothesis directly. So that's another reason we might want to have in the back of our minds that this could be useful.
Allison [00:11:07] Absolutely. I think there's another question we can answer too, which may be a little less important, but not necessarily less interesting to interested parties. And that's you know, I think, you know, we've all heard talk of the idea that this particular population may be particularly partisan in one way or the other. Right. And so finding out about this population, if we can get a list of these people, understand how representative it is of the larger population, we can also get a sense of whether or not these assumptions are true.
Jennifer [00:11:40] Yeah, as we will talk about, it's been we know remarkably little about this, obviously.
Allison [00:11:45] Really.
Jennifer [00:11:46] Yes. Okay. So in this paper, we're going to try to find people with felony convictions who are eligible to vote in their home state and encourage them to register to vote. So before this study, what had we known about how to increase the civic engagement of this group in particular and of people in general?
Allison [00:12:02] So this population, I think when I think of what we knew about registering this population, I think of two things in particular. And so one finding that was really important was that you can register these people to vote with, you know, the kinds of interventions that we did in our project, but sort of what this previous research did is, is that it relied on a list provided from the state. And so it was really helpful and I think laying the groundwork for the kind of work that we do here, but prompted us, if I can speak, for all of us, to try to think bigger. Right. So in what ways was the list of individuals from the state narrow, in what ways, you know, was it not representative of the larger population. And in getting a more representative list, can we actually, quote unquote, do better, get more people registered, for example.
Allison [00:12:55] Another, I think, particularly important thing we learned from previous literature in this area is that the knowledge gap really does exist, right. So knowing whether or not you are eligible to register was actually a real barrier. I think previous work found this with a study in Maine. No, Massachusetts. It started with an "M", I'm pretty sure, but that sort of letting people know that hey, in fact you are eligible had a huge impact. Right. Or had a real impact on whether or not people were registering to vote. So I think these are two sort of really important findings that lay the groundwork for the kind of work that we're doing in our project or projects so that's really important.
Allison [00:13:35] And then I think more broadly, what it can help, you know, what we sort of gather from existing research is that one much of existing research on mobilizing people, getting them out to vote is really focused on that voting piece, right. So what we're trying to learn from in this study comes from studies that are doing a slightly different thing, right. So not reincorporating people, but getting those who are already registered to vote and what's nice about that is that those studies had lists of registered voters, but what that doesn't necessarily help us with is creating that list of people we might potentially register. And so the other thing that we can learn from sort of existing research is trying to figure out how to think of returning citizens are they just unlikely voters, are they more like new voters so people who have just aged into voting, for example. And so trying to sort of parse that out, I think also helps us think about how best to characterize the individuals that we're trying to register here.
Jennifer [00:14:44] Yeah the piece about how most of the research before this had been focused on how to get registered voters to turnout, so get out the vote efforts was really surprising to me, actually.
Allison [00:14:55] Oh really.
Jennifer [00:14:55] Coming to this literature as of for the first time, I sort of had this impression that we know a lot there have been all these, you know, RCTs done of like how to get people to register to vote. It turns out we know a lot about how to get registered voters to turn out.
Allison [00:15:09] Yeah.
Jennifer [00:15:10] There's still more to do, but we know a lot about that, but we essentially know nothing about how to broaden the electorate. Am I characterizing that correctly?
Allison [00:15:19] Yeah, no, we really don't and it's like.
Jennifer [00:15:20] Which was just sort of yeah, surprising.
Allison [00:15:23] Yeah. It's like, how do you find people when like so one thing that I won't even say, like easy helpful for us is that we can at least get a list from the states of individuals who've been in the system. Right. Okay. So that makes that our a difficult process a little bit easier, but what if we were trying to figure out how to find other unregistered voters? How do you even begin? What administrative list would you start with there? What other kinds, you know, ways can we sort of attempt to go about this process? And I can imagine it getting even more and more difficult. Right. So, yes, growing the electorate is really hard because you can't find what you don't know how to look for.
Jennifer [00:16:03] Yeah. And so this leads into my next question about like why we don't actually do so. So there's sort of this quest, this issue that like in the United States, there is no centralized list of all.
Allison [00:16:14] Of all the people.
Jennifer [00:16:16] Right. Of all the people that might be eligible and so you have to start somewhere. You have to start so people that have thought about this, I guess, start with like this, a student with college students or something. So, yeah, so say more about that, about like the data challenges and then and then identification challenges like what? You know, even beyond getting the best data. How do we tell what actually works, what's effective and why is that hard?
Allison [00:16:42] Oh, gosh, so many reasons. Okay, so first of all, in terms of figuring out who it is we're targeting, this requires the merging of multiple different datasets and in different ways. So first, we need to figure out in a given state who are the individuals with previous felony convictions who are now eligible to vote, and then how can we contact them? What are their addresses? So we first start off with a list of names of individuals who've been in a State's Department of Corrections. And so sometimes that is publicly available, relatively easily downloadable information and so you can gather that information from the website or requested from the state. And it just tells you who's been in the Department of Corrections in different places or whether they've been in whether they have been or are incarcerated, probation, parole, etc., and then from there, we need to identify which individuals are eligible based on a state's rules. So let's say those rules mean they have to have completed any sentence of incarceration and related parole that follows it. So then we reduce that list to those individuals and then we do some further reductions.
Allison [00:17:52] So we might do some reductions based on age like we did to allow individuals who are more likely to be deceased at this point, for example, and kind of narrow it down to these are the set of people that we think are eligible to vote because they finished out their sentences that are likely to be alive. We can do some checks to make sure they haven't thought of they haven't been convicted again since that point. Right. And so we get this list of eligible people we think are eligible voters. Then we have to pull down the state's voter file that tells us who in that state is registered to vote. And what we want to do then is take our list of eligible individuals with former felony convictions and pull out anybody who is already registered since we can't register somebody who's already registered.
Allison [00:18:42] So now we've got a list of individuals with past felony convictions who are eligible to vote, who are not currently registered to vote and then we need to find them. And so here's where we had to partner with a commercial data vendor and see if we can get what seem to be good addresses for these individuals and remember, this is a you know, I have to tell you to remember, but to anyone listening, right this is a relatively transient population. And so we have to realize that there may be a number of people for whom we can't find good addresses and people, you know, who, when we do find good addresses like may not be at those addresses anymore. You know, the reentry process is incredibly difficult and has so many challenges. So, you know, it can be hard to find people even we feel like we have good data, right. And so from there, we have a list of people who are eligible voters who are not currently registered to vote and for whom we have addresses.
Allison [00:19:37] And so identification challenges, right so we need to figure out the kind of messaging that we think will be strong enough to have the potential to get these people out to vote and also the delivery strategy I mean, we used mailers we mailed people pretty simple mailers, I guess we'll get into in a bit more detail, too, but letting them know what makes somebody eligible to vote, letting them know the election was coming up. All of that sort of information and oftentimes providing them with the sort of registration packet. So registration form, addressed envelope with a stamp and so we made it relatively easy for most people.
Allison [00:20:13] We also tried out some other strategies of communication, of reaching people we tried out text messages, we found out those weren't particularly effective. I believe there were some phone calls made, but what we landed on was that through a series of pilots that if we're going to see an effect, this is actually the way to go, sending out these mailers.
Jennifer [00:20:33] Yeah, and it might be because our address data was relatively good, which I think was sort of a surprise to most of us.
Allison [00:20:39] It was to me.
Jennifer [00:20:40] Yeah, it worked as well as it did the commercial vendor had pretty good data. Yeah. So just on the a little bit more on the identification piece, so yeah so I think in general, you know, you want to have some sort of control group and.
Allison [00:20:53] Oh yes all that stuff.
Jennifer [00:20:54] Like this. And so yeah, so we're going to run a randomized controlled trial. I think that's actually fairly typical in this space. There are a lot of, especially for the GOTV experiments. So this is a space where maybe we worry less about identification than in some other situations.
Allison [00:21:11] Yeah, we used a pretty sort of classic approach based on things we gathered from previous research. So simple mailers, we had five conditions, so five experimental conditions, one pure control. So these were individuals that we didn't contact at all, they're on our list, they're eligible, they're not registered, we have valid addresses and this is our pure control group, they're not getting any mailers. And then we had four variations on our mailer that we sent out. And so across our five groups, so control and then the four groups received letters individuals were randomly assigned to each group, so with equal probability of landing in each group. And so I believe our four letter categories were, let's see.
Allison [00:21:53] So we had this basic mailer packet that I kind of briefly mentioned before that includes information about the requirements to be eligible to register and vote in the state, includes an address and stamped envelope as well as the registration form. It also includes some language about about being eligible if you had a past conviction. And then we have a letter that is the same as that, but removes the information, the language about voting when you when you've had a previous conviction, then.
Jennifer [00:22:27] It's still in the list of eligibility requirements.
Allison [00:22:29] Yes, it's in there.
Jennifer [00:22:30] But it's just not like highlight.
Allison [00:22:31] Exactly. Exactly. And then we have one that is the same letter as the original one ao it still has that language highlighting that you can vote even if you had a conviction, but it does not include the envelope and registration form. And then we have a final one that is the same as the original full mailer packet, but it includes civil rights framing. So, you know, you want to vote, you want to make sure you have all of your rights that kind of language that research has shown, you know, may encourage people to exercise their rights. And so those were sort of that's the outline of our five different conditions that include four different mailers.
Jennifer [00:23:09] So what was the motivation for that kind of highlighting that language at the beginning in our basic mailer and then in that extra civil rights reading mailer for that kind of, you know, your rights are on the ballot.
Allison [00:23:21] Right right.
Jennifer [00:23:22] Criminal justice matters. Why'd do we do that?
Allison [00:23:25] Well, basically, we wanted to see if appealing to the kind of sentiments we think might drive this particular population would make a difference. And previous studies that, you know, not focused on this population, but like get out the vote studies that have included information like exercise your right if you want to be heard, you need to you need to get out and vote have you had some level of effectiveness and so this is what we're tapping into here. I think another reason we did it is that, you know, we partnered with a local or a state nonprofit organization that's really interested in mobilizing this particular population. And they also felt very strongly that it was important to have the kind of language upfront that let people know, hey, even if you've had a conviction, you can vote as long as you're off papers, right? That's the language that they that we ended up using.
Allison [00:24:16] And what was interesting, I think, to me as a researcher, I haven't had as much experience partnering with nonprofits in this way. I have in terms of collecting data and going is sort of like sharing information with them, but not partnering in this way where we were reaching out to citizens together. And so what was really interesting to me as a researcher was being involved in these conversations where members of the nonprofit were letting us know what they know works, right and they've worked in this space for really, really, really long times. You have to respect that. Then as a researcher, knowing that, well, you can't really know because you've never tested it. Right. And so this was, I think, an important learning moment for me, because you really do have to be very respectful of people working in this space because they have a lot of knowledge that we don't have as researchers. And they're also helping us out, right, because they're partner with us in research so we can make our research really strong. And I like to think we can help them out too, by getting them access to the kind of information they wouldn't otherwise, which in this case is potentially letting them know if this language that they truly believe is important actually is important. Right. Because, you know, well, you can't know if you've never tested it, which is, you know, our hill, I think, is social scientists.
Jennifer [00:25:34] Right. Right. Okay, let's see. Let's go back to the list briefly. So what did the people on our list look like? So I guess we should say we probably mentioned at some point this is North Carolina that we're focusing on. Actually, let's back up even one more step. So in this focus, we're in this paper, we're focusing on people in North Carolina in the lead up to the 2020 general election.
Allison [00:25:56] Yes.
Jennifer [00:25:56] So what were the state laws governing voter registration in that state at that time?
Allison [00:26:00] Right. So the state laws governing registration were I mean, you have to be 18, you had to live in the county you lived in for I don't remember if it was a full year or a month, but important for this project was that you had to have completed your entire sentence if you'd been convicted of a felony, including all incarceration, all parole. So as long as you'd finish those portions of your sentence and you were 18 and met all of the other requirements to register, you were eligible to register and vote.
Jennifer [00:26:26] Yes. And you could also register online.
Allison [00:26:29] Yes, that's right.
Jennifer [00:26:30] Which is nice. Yeah. So next. What do the people on our list look like? So we go and get this the criminal history data from the Department of Public Safety and merge it, as you said, with the voter file and then get the address data.
Allison [00:26:45] Yes.
Jennifer [00:26:46] So who winds up on our list? What do they look like and how representative are they of the broader population we were targeting?
Allison [00:26:52] Yes, absolutely. So the good news for us is that the people on this list look really similar to the broader population of individuals we're interested in reaching so that's people who have passed on convictions and who may now be eligible to register and vote because they finish their sentences so overall, they were really similar the populations both were about 75% male. I think 76% in the full population and 75% in our list of people with addresses so really similar. So our list and then the broader population that comes from it was about just over 50% black individuals. And the one place that our list kind of differed in any significant way from the broader population was that it was a bit younger. This is most likely because we cut or we dropped from our list anybody who was 70 or above and so it makes sense that our list would be a little bit younger. I think the average age in our list is 42, and the average age in the larger list was something like 50. And because of that, also people in our list had had a shorter time since released on average than the overall population, but again, our list was younger and didn't include people who were maybe like 80 years old and, you know, had their conviction, you know, three decades ago or something like that.
Jennifer [00:28:12] And to be clear, so we made that that's that age cutoff we cut off folks who were 70 or older because we were being conservative.
Allison [00:28:20] Yes.
Jennifer [00:28:20] In terms of wanting to err toward making sure people on our lists were still alive.
Allison [00:28:25] Absolutely.
Jennifer [00:28:25] And we didn't have the death information.
Allison [00:28:27] Absolutely. Yeah. I mean, there's a world in which we could have, you know, cut it much later. But because we didn't have that information, we just wanted to be extra sure.
Jennifer [00:28:34] Yeah. And then so how long had they been out on average?
Allison [00:28:38] So in the full list so in the full sort of population of interest, the average time since release was 17 years and ours it was 9 years.
Jennifer [00:28:47] So they've still been out a fair.
Allison [00:28:48] Oh yeah.
Jennifer [00:28:49] Amount of time, which is also really striking basically the one other study out there done with this population that you alluded to earlier where basically they got a list of sort of a narrow subset of folks that were recently released on.
Allison [00:29:02] Yes.
Jennifer [00:29:02] Parole, I think essentially in Connecticut.
Allison [00:29:04] That's right.
Jennifer [00:29:04] Those folks were, you know, had just gotten out, basically and so we're going to be targeting people that have been out in their communities for many years.
Allison [00:29:13] Absolutely. Yeah. I think that's another thing that's exciting about it. Right. Imagine somebody been released for a long time, you know, hopefully sort of like fully reintegrated into society. They have their routine, they have their lives maybe they still don't know that they're eligible to vote, maybe they haven't thought about it, maybe nobody's ever reached out to them. You know, I think it's exciting to think about being able to reach out to that population, especially.
Jennifer [00:29:34] Yeah. Okay. So we we have these mailers now that you that you told us about, we're going to run this is as a randomized experiment. We've got our our list of people. We randomly assigned them to a control group or one of these four mailers. What outcome measures are we interested in to measure the effects on voter participation?
Allison [00:29:52] Yes. So we're interested in two outcome measures. Our main outcome measure of interest is whether or not people register to vote. Our secondary outcome of interest is whether or not they actually turn out to vote once they're registered.
Jennifer [00:30:06] Yeah. And this is it may it's helpful because this is public information, it turns out.
Allison [00:30:10] Yes.
Jennifer [00:30:11] Which is amazing out there. Yeah on who register, who's registered in the state and who actually voted in the election. We don't know how you voted, but we know if you voted.
Allison [00:30:20] Exactly. And that's that's really all that matters for us at this point.
Jennifer [00:30:24] Right. Exactly. Okay. Well, let's talk about what we found. What were the effects of our mailers on voter registration and turnout in November 2020?
Allison [00:30:33] So overall, the strongest effect of our mailers was on the outcome we were most interested in, which is registering to vote. So what we found was that receiving a mailer so regardless of the type of mailer that people received, receiving a mailer compared to the control condition where they received no mailer led to a 0.8 percentage point increase in registering to vote. This translates to about a 12% increase, which is a pretty big effect.
Jennifer [00:30:59] Yeah, that 0.8% sounds small, but super low baseline.
Allison [00:31:03] Exactly. Exactly right. Compared to people who aren't registering like this is a really big deal.
Jennifer [00:31:08] Mm hmm. Yeah. And it was also it's also been interesting to talk to folks who work in the practitioner space here, and they hear about this 0.8 percentage point effect and they're like, whoa, that's huge.
Allison [00:31:19] Yeah.
Jennifer [00:31:20] Which is very gratifying, especially in a general election like this, where everybody's flooded with information.
Allison [00:31:25] Yeah.
Jennifer [00:31:26] We were still able to reach this group that apparently no one else was reaching.
Allison [00:31:30] That's true. Right. Like, think back to, you know what's going on in 2020.
Jennifer [00:31:37] Mhm.
Allison [00:31:37] Right. Like this.
Jennifer [00:31:38] Yeah.
Allison [00:31:39] A lot is happening.
Jennifer [00:31:39] Major presidential.
Allison [00:31:40] Exactly.
Jennifer [00:31:41] Yeah. Yeah. And North Carolina. Swing state.
Allison [00:31:44] Yeah. So to make any waves there I think is you know like that's really encouraging to make any waves and sort of that environment.
Jennifer [00:31:50] Mhm.
Allison [00:31:51] So in terms of you know whether or not our mailers made people vote, so the effects on whether or not an individual actually voted were a bit more modest. So 0.5 percentage points which translates to about 11%, you know, marginally statistically significant here. We're not talking about the same level of certainty that we might have in our affects on registering to vote.
Jennifer [00:32:13] Right. Still a big effect, but yet we get smaller as we or less significant. Great. Okay. So that's the that's the big punch line.
Allison [00:32:21] Yes.
Jennifer [00:32:21] Is that this sort of low touch mailer sent to these people that apparently no one else was targeting actually got them to register to vote, at least, you know, about a 12% increase. We preregistered as we preregistered a whole bunch of stuff.
Allison [00:32:35] Yes.
Jennifer [00:32:35] Which we preregistered a particular interest in how these effects differed across racial groups. So what did we find when we compared effects for white and black recipients of our mailers?
Allison [00:32:44] So we were actually pretty surprised when we sort of looked at the results of our heterogeneity analysis and, you know, tried to figure out what was going on here in terms of whether or not there were any differences in our effects across racial groups and we didn't see a lot. In fact, we saw that, if anything, our mailers were most effective, most effective among white returning citizens.
Jennifer [00:33:09] Yeah, which is definitely not what we'd expected.
Allison [00:33:12] That's not what we expected.
Jennifer [00:33:13] Yeah, we and we did this a couple of ways. I think we like split it white and black.
Allison [00:33:17] Yes.
Jennifer [00:33:17] And basically almost everything's driven by the white recipients. And then I think that difference is not statistically significant, but it's pretty darn suggestive like.
Allison [00:33:26] Yeah.
Jennifer [00:33:28] We just don't quite have enough power.
Allison [00:33:29] Exactly. We don't have enough power, but it's also just not what we would have expected directionally.
Jennifer [00:33:34] Right. Especially given kind of the general conventional wisdom about this group that it is you know, we know that people with criminal justice contact are disproportionately people of color, black and brown, disproportionately black men in particular and so it was striking that we see a bigger effect of our intervention on white people. Of course, there are still lots of white people who have felony convictions, so.
Allison [00:34:01] Exactly. Exactly.
Jennifer [00:34:02] Yeah. And then we also did this fancy machine learning thing for heterogeneity, this heterogeneity on other other other dimensions kind of like throwing everything else we knew about people in there.
Allison [00:34:14] This was your big thing, Jen. I feel like you should talk about this one.
Jennifer [00:34:17] Following some other other economists, especially Sara Heller and Jonathan Davis, used this in their paper on summer jobs programs and trying to figure out who benefited the most from summer jobs. So we we took that essentially off the shelf, initially designed, I should say, by Athey and Wagner, I think and so we we did this and threw a whole bunch of stuff in there to see if there was anything else interesting that that predicted which groups might have bigger treatment effects and what did we find there.
Allison [00:34:49] So that machine learning approach sort of confirmed what we saw before in terms of the racial group where we saw sort of the biggest effect. And the other things that came out of that were biggest effect among individuals who are male right, which isn't too surprising. And also those who'd been in prisons and not just in probation or, you know, some other type of supervision, but people who had been incarcerated.
Jennifer [00:35:10] Yeah. And no other clear patterns on other stuff.
Allison [00:35:14] Yeah.
Jennifer [00:35:14] That we thought that was or their age or time since release, which I think I was sort of surprised by the time since release wasn't a clear a predictor.
Allison [00:35:21] Yeah.
Jennifer [00:35:21] Yeah. And then were there any other interesting differences in the effects across different mailer types since we had all these different arms for mailers?
Allison [00:35:29] Perhaps the most interesting thing is that we actually didn't see much difference across the other types of crime.
Jennifer [00:35:34] Right.
Allison [00:35:35] This is again an area where we were underpowered. So, you know, future work could potentially like just scale up and actually see if there's any real difference across these types of mailers.
Jennifer [00:35:46] Well, I mean, we were underpowered to find small effects, so it's possible that there were small effects, but I think we kind of went into it thinking that there would be big effects across the language, across these different mailers.
Allison [00:35:59] Or at least likely including the form and the envelope, but I mean, the thing that could have made a difference there, right, was that they could register online.
Jennifer [00:36:07] Right.
Allison [00:36:07] So maybe in a world where there is no online registration, we see a huge effect from having the registration form and the address stamp envelope included.
Jennifer [00:36:17] Right yep. Okay. And then we ran a few other tests to try to understand our results a bit better. One thing we wondered was whether the different effects across racial groups might be due to differential address quality. So maybe the address data we had for black recipients was just les accurate then the address that we had for white recipients and so our mailers never even reach them and so that's why we we saw such small effects for black people versus white people. So we sent a follow up postcards to test for this. What did we do there and what should we find?
Allison [00:36:48] Yeah, we sent follow up postcards to people trying to see if basically if these postcards landed and then giving them an opportunity to, like, do a little survey, I think, and get what was it like an Amazon gift card or something? And we found that we didn't find any real differences in terms of address quality based on recipient race, which was so basically no support for that hypothesis in terms of why we see these different effect sizes and stronger effect sizes among white recipients of our mailers.
Jennifer [00:37:16] Yeah, and I don't think we actually did anything with the survey data, right. We were just sort of like, well, while we're sending them postcards to check their address quality to see if it bounces we might as well try this, but.
Allison [00:37:27] We didn't I had actually even forgotten that we had the survey piece until I was saying this thing to you now, about the postcards, it's like, oh, there was this survey.
Jennifer [00:37:36] This is one of those like a research like sausage making where it's like, well, we're sending them this postcard we might as well put something on it and then like, but I think none of us really expected the response rate to be high.
Allison [00:37:47] I mean, like always collect more data rather than less data.
Jennifer [00:37:50] Exactly. Exactly. I guess we should go back and see if anyone responded to our survey. Okay. And then finally, we also ran a parallel experiment where we sent mailers to a comparison group of people without criminal records who lived in the same area as our main target population. So why did we do this and what did we find there?
Allison [00:38:09] Yeah, so what we wanted to see was whether or not something about our mailer was particularly effective among this population we're interested in returning citizens who are eligible to vote. And so, as you said, we found we created a list of individuals who were from the same communities as people who were on our list, but who, as far as we could tell, had no past felony convictions. And they all got our basic mailer package minus the extra highlighted criminal justice framing and we actually don't see any effect of our mailers among this population. And so what does this mean? We're not entirely sure, but it does suggest that there may be something especially effective about reaching out to the population that we reached out to, or perhaps that they are more easily mobilized than others who have not had their rights taken away but still remain unregistered for example. We don't really know, but I do think it's a really interesting piece of evidence.
Jennifer [00:39:12] Yeah, so it does yeah, it kind of gets this thing that we were talking about earlier where like, did we just do an experiment trying to reach economically disadvantaged people? And this is the effect of mailers on economically disadvantaged people and so this comparison and parallel experiment says no, right there's something different about this group.
Allison [00:39:31] Right like, it doesn't tell us exactly what that different thing is that we've actually tapped into, but that it is different rhat it is particular.
Jennifer [00:39:39] Mm hmm.
Allison [00:39:40] Which I think is helpful because we're interested in this population.
Jennifer [00:39:43] Right for sure. And so, yeah, when I think about this, I think the possibilities kind of fall into two main buckets. Like one of them is we're reaching a population that others are not reaching.
Allison [00:39:55] Right.
Jennifer [00:39:55] So our list is just fundamentally different from the lists that other organizations are trying to contact that are working in this area or the information that we're sending them is uniquely valuable to this group. So we're telling them if you've a criminal record, you can still vote. And maybe other people aren't saying that.
Allison [00:40:14] Right.
Jennifer [00:40:14] And so or could be both of those things. But yeah, it was at least reassuring that we saw such a stark difference and it really is like I mean, for the for the comparison group, it's just it's not like an imprecise null it's like 0.
Allison [00:40:29] Exactly.
Jennifer [00:40:30] There's 0 effect on that group. They are not at all affected by our mailer.
Allison [00:40:34] Exactly. And I was really excited to see that, like, there's something we're doing that's really getting to this population that.
Jennifer [00:40:40] Yeah.
Allison [00:40:40] You know, remains largely unregistered. And I'm really excited to keep trying to figure out what it is.
Jennifer [00:40:48] Yeah.
Allison [00:40:48] But it's a super exciting result.
Jennifer [00:40:49] Yeah.
Allison [00:40:50] Even if we don't have the answers around it, you know, like it's still confusing.
Jennifer [00:40:54] Right in some ways it raises more questions.
Allison [00:40:56] Exactly.
Jennifer [00:40:56] Um I guess we could also mention, I mean, we we've since been working with another organization that does a lot of outreach to try to broaden the electorate. And we actually compared our list with theirs. It might have been in another state, but but basically there was really very little overlap like they were I think everyone was surprised how unique our list was, which really says something about how difficult it is to find the contact information of people who have criminal justice involvement for some reason. It is difficult to get them onto other lists that are made.
Allison [00:41:29] Exactly. And there's like there's a number of organizations out there who are interested in mobilizing populations that are similarly unincorporated into the electorate, for example. And so we don't know we don't know how each of these organizations is pulling their lists together, whether it's in any systematic way. It's probably in many cases, right just like meeting people and spreading out and going to events in the community and things like that. And so, you know, hopefully we've created a way to to do this that helps to sort of expand the reach of efforts trying to reach this population and and let them know about their their rights.
Jennifer [00:42:10] Yeah. We should also mention that we looked at what party people registered as.
Allison [00:42:17] We did.
Jennifer [00:42:17] So we couldn't see how people voted, but we could see if they registered as a part of a party when they registered to vote, we could see how they self-identified. What did we find there?
Allison [00:42:27] Yeah, we didn't really find much of a difference there. We had nearly equal proportions of Democratic registrants and Republican registrants. So at least in our study, right. We did not find that this population was particularly democratic, which is usually the assumption.
Jennifer [00:42:46] Right. So at least in North Carolina.
Allison [00:42:48] In North Carolina.
Jennifer [00:42:48] Then also relative to our control group and our comparison groups like it, it wasn't like our we also found that our intervention wasn't especially good at registering people from one group or another.
Allison [00:42:59] Exactly.
Jennifer [00:42:59] Which was also nice because this was we should highlight a nonpartisan effort. We were not going into this trying to register people of a particular political leaning, but I think our prior was, as it is for most people, that this would be a more left leaning group just because that's the way it's talked about in the news and it wasn't.
Allison [00:43:20] Right.
Jennifer [00:43:20] Which I thought was great. You know, it's like, this doesn't need to be a partisan fight, right. It doesn't need to be a political football whether we let this group vote or not or try to encourage this group to vote or not, we could just do it because it's good for democracy.
Allison [00:43:36] Yeah, And I think, you know, another I think one reason people have the assumption that this group will be particularly democratic is because black and brown people are overrepresented in this population, but even though they are overrepresented, it doesn't necessarily mean they're going to be the vast majority of the population.
Jennifer [00:43:53] Right.
Allison [00:43:53] And so this is not to say that we're not going to have white individuals who are registered as both Republican and Democrat, but I think, you know, there is this assumption that it's going to be like 80% black and brown people in this population, but that's just not true.
Jennifer [00:44:08] Right.
Allison [00:44:08] Right. And so I think thinking about it in that way should sort of temper people's assumptions a little bit about what you're going to find when it comes to partizan identification in this in this population.
Jennifer [00:44:19] Okay. So what are the policy implications of all of this? What should policymakers and practitioners take away from these results?
Allison [00:44:26] I think practitioners especially should take away that it's possible to put together a list of individuals in this population for whom they can find good quality addresses. So that's one thing. And inasmuch as they have the resources to do what we did in this situation, they can probably really broaden the scope of people they're able to reach and potentially register and potentially get out to the polls. So that's one thing. And it might help to sort of increase the efficiency of their practices as well. Right. So if we can continue to figure out the kind of interventions that work. So the kind of contact that works with this population and we find that sending out pretty sort of low touch, as you said, letters with relevant information, maybe not even needing to include a registration form, an envelope, you know, that's cost savings there that they might be able to then put into sort of reaching a larger swath of this swath of this population.
Allison [00:45:24] So, oops. So I think that's something that could really be useful to two practitioners. Now to people working in policy one thing that I think would be really helpful and as much as these individuals sort of want to be helpful here, is that letting people know they're eligible to vote can be a really important piece of this puzzle. I think it was and I will have to double check this and let you know for sure, but I think it was a piece by Meredith and Morse, right that found that, you know, a big hurdle here was that people weren't aware that they were eligible.
Allison [00:45:57] And I think in the study, they randomly assigned individuals who were being released to receive a letter letting them know that they were now eligible to vote again and they saw effects from that. Right. And so this is one place that policy practitioners can comment that people in policy spaces can come into play when people are discharged, let them know they can vote or perhaps like include a registration for in their discharge materials.
Jennifer [00:46:24] Yeah, we wound up having a lot of conversations, especially over Twitter, I think about why we, you know, why we were running this experiment instead of just like standing outside of a prison with a clipboard and registered to vote there and part of it as a matter of scale and and we can reach more people this way and we want to know if it worked. And part of it is kind of the timing of, you know, when someone walks out of prison, they're not necessarily eligible in that moment.
Allison [00:46:49] Right.
Jennifer [00:46:49] So it might. So to some extent,.
Allison [00:46:51] Exactly.
Jennifer [00:46:51] This could be a question of like, when do you tell them, you know, do you tell them when their parole is wrapping up.
Allison [00:46:56] Exactly.
Jennifer [00:46:57] Or their probation is wrapping up, but yeah, it definitely I mean, there is now this stock of people who have been out of prison for a long time that we need to find ways to contact, which is what we were focused on, but there's also a flow of people coming out every day that we could be doing a better job of informing at the time when they can use that information.
Allison [00:47:17] Exactly. So maybe when they leave prison, maybe it's in their wrapping up in parole or probation, but this seems to be something that wouldn't be that difficult to implement. I mean, people are getting information when their sentences are wrapped up anyway.
Jennifer [00:47:29] Yup.
Allison [00:47:29] This would be another piece of information.
Jennifer [00:47:31] Yeah. And it's, you know, in some ways an open question about like, would it be as effective to tell someone when they're first trying to reintegrate and trying to get a job and everything else? Maybe this is not maybe they just forget about this other piece of paper and so actually contacting them a year would be better, but to the extent that this is mostly an information problem, telling people repeatedly at every opportunity would be helpful.
Allison [00:47:57] Yeah. So if we're going to disenfranchize people at all, which I think is another question, at least let them know when when they can access that franchise again. Exactly. Exactly.
Jennifer [00:48:08] Yeah. I mean, I think the other big implication for me of our results is that, you know, even this, again, super low touch mailer, like I almost think of this as like a proof of concept kind of experimen like, you know, you could imagine a much more involved intervention where you're like knocking on doors and actually talking to people or whatever, but even the low touch mailer was super effective, which suggests that like that hypothesis that was floating around or that that some people you know, the the the belief that this group is just is voting at low rates because they just they're not interested, they don't want to be involved. That hypothesis really, you know, I'm sure it's true for some people, but we found pretty big effects for our mailer, you know, on a level that how do I want to say this, we found effects from our basic mailer at a level that that was, you know, pretty big relative to what we would find a mailer for other populations.
Allison [00:49:06] Right.
Jennifer [00:49:06] And so it suggests that, like, these people were just sort of waiting to be.
Allison [00:49:10] Yes.
Jennifer [00:49:10] Asked or waiting for the information, which I think a lot of groups out there could run with.
Allison [00:49:15] Yeah, I agree.
Jennifer [00:49:17] Yeah.
Allison [00:49:17] I agree. I think that's that's the really exciting part.
Jennifer [00:49:20] Yeah. So Desmond Ang and Jonathan Tebes have a paper looking at the civic responses to police violence. And so they look at what happens to voting behavior of people who live near a location where there was where someone was killed by police.
Allison [00:49:35] Okay.
Jennifer [00:49:37] And they find that registration and voting increases and those effects are driven entirely by black and Hispanic residents and are largest in incidents where it seems like it was a more unjustified killing.
Allison [00:49:50] That makes me think of sort of like the broad thesis of our coauthors book Hannah's book.
Jennifer [00:49:54] Exactly.
Allison [00:49:55] Yes.
Jennifer [00:49:55] It's very much in line with Hannah Walker's "Mobilized By Injustice" and we interviewed her or David Eil interviewed her for the podcast a while back, so I'll put a link to that in the show notes. But yeah, this idea again, that that's very counter to this other narrative that people who've had this negative interaction with the criminal justice system now are like lost the polity and they don't want to engage anymore. This work, both this paper and Hannah's work, suggests that actually having these negative interactions moves people to be more engaged.
Allison [00:50:24] Yeah.
Jennifer [00:50:25] And so which is again, it's another piece, another, you know, piece pushing us toward this idea that they're just kind of waiting to be out there, waiting to find out if they can, which is just fascinating to me.
Allison [00:50:37] Well, I mean, this I mean, this points to another sort of takeaway from this research that I forgot to mention before, but like, this is really useful information to campaign for campaigns, right. Campaigns that really want to spend their dollars where they think they will be most effective and are largely ignoring this population, it seems r.ight, because they assumed because one, they don't find them on the voter registration list on the on the registered voter lists. And two, they probably assume that they won't be able to be mobilized or, you know, this is an untapped area for them. These are people who at least, you know, we find in our study so far seem to be, as you said, ready to participate, ready to register, ready to vote, these are new voters, right.
Jennifer [00:51:22] Yeah.
Allison [00:51:22] So you could bring them in. And this is something that could be really useful to campaigns as they think about their strategies, especially if it can be done, you know, through something low cost like a letter.
Jennifer [00:51:32] Yeah, so what's the research frontier? What are the next big questions in this area that you and our team and other people are going to be thinking about going forward?
Allison [00:51:42] Yeah, so I think a few things. For one thing, seeing if we can find out more about the differences, the potential differences in our effects across racial groups. Two seeing if we can identify what it is specifically about either our letter or this population that made it particularly effective among returning citizens and not among our similar group of individuals who had not had prior felony convictions so those are two things. Three, seeing how this kind of setup works in other contexts. So other states with maybe different registration set ups, different rules, and then some, you know, sort of fun ways that we're looking to expand that are a little sort of further afield from this original experiment are one thing that we're excited about is seeing whether or not we can mobilize people through, you know, family members, household members, things like that. So, you know, maybe we reach out to the individual, but maybe we also reach out to somebody who lives in the house with them, somebody who's already a voter and that person can really encourage them to get out and vote.
Allison [00:52:53] So that's one sort of direction that we're already looking into. And another one and this is probably for like farther down the line. Are there other populations that are maybe similar to returning citizens and that, you know, they typically don't participate, they're typically sort of not frequent voters, maybe more likely to be unregistered, that we can use administrative data to find and contact in similar ways and what would those populations be.
Jennifer [00:53:27] Yes, you alluded to this, but we are we are we are still working on that. We just ran an experiment in 2022 in Texas and a little bit of follow up in North Carolina. So stay tuned for those results. Yeah, and I'll just add something that I mentioned near the beginning that I think I'm all I'm interested in kind of how these pushes to engage with the community or vote become more involved in the democratic process, how all of this affects people's longer run outcomes, how it affects their reintegration to society after after jail or prison or other other kinds of outcomes in the community. And so I think there's definitely a lot of work to be done there, trying to see like what are the effects of either policy changes or increases in in voter engagement on a variety of other outcomes.
Allison [00:54:18] Definitely. Definitely.
Jennifer [00:54:20] Well, my guest today has been Allison Harris from Yale University. Allison, thank you so much for talking with me.
Allison [00:54:24] Thank you for having me. This was great.
Jennifer [00:54:31] 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 Patreon 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.