Probable Causation

View Original

Episode 40: Emily Leslie

Jennifer Doleac

Emily Leslie

Emily Leslie is an Assistant Professor of Economics at Brigham Young University.

Date: November 10, 2020

Bonus segment on Professor Leslie’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. Leslie's work on how COVID-19 has affected domestic violence:

"Sheltering in place and domestic violence: Evidence from calls for service during COVID-19" by Emily Leslie and Riley Wilson.

OTHER RESEARCH WE DISCUSS IN THIS EPISODE:

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 Emily Leslie. Emily is an Assistant Professor of Economics at Brigham Young University.

 

Jennifer [00:00:24] Emily, welcome to the show.

 

Emily [00:00:26] Thanks, great to be here.

 

Jennifer [00:00:28] Today we're going to talk about your research on how the covid lockdown has affected domestic violence. But before we get into that, could you tell us about your research expertize and how you became interested in this topic?

 

Emily [00:00:39] So my research has focused on crime in the criminal justice system for a while. My first published crime paper started out as an undergraduate term paper. So I'm kind of in the mix with my research interests kind of from the beginning. As for this specific paper, Riley and I had recently been looking at the data repository that we use for our main data source and thinking for a few months about what projects we could use it for. We were sure there were some interesting ones in there. And then covid hit and we started to see articles coming out about domestic violence during the pandemic. And we said, well, let's see what we can say about it using this data on police calls for service that we are already interested in using.

 

Jennifer [00:01:25] Yeah. So as you note, almost as soon as social distancing and stay at home orders happened this spring, people started worrying about whether domestic violence might increase. So walk us through why this is. What are the reasons that these changes in behavior might have caused more domestic violence?

 

Emily [00:01:40] I think the thing that's probably top of mind for most people is just the sheer number of people who are suddenly spending way more time at home, both at of, you know, by choice and because of shelter in place policies that were limiting their ability to gather with other people.

 

[00:01:55] And all of that is in the context of a pile of new stresses. You have, you know, about 35% of workers switched to remote work. And those are the lucky ones because there's also a 10 percentage point increase in unemployment from January to April. So there's just massive labor market shocks and you've got schools closing and going online and then the pandemic itself. So there's layers upon layers of stress on top of the fact that people were kind of forced together more. And I think that for many observers, that was intuitively a dangerous situation, one that could be prone to encouraging higher rates of domestic violence.

 

Jennifer [00:02:37] So your paper is titled "Sheltering in place and domestic violence Evidence from calls for service during COVID-19," and it's coauthored with Riley Wilson. So in part, this paper is simply measuring the effects of recent policy decisions and the ongoing covid crisis on one important form of violent crime. But your paper also tells us something about the underlying causes of domestic violence. So before this study, what did we know about when and why domestic violence increases?

 

Emily [00:03:04] There are a few great studies out there about domestic violence. It can be a difficult topic to study, but there are several papers that tell us that there is definitely a connection between labor market conditions and domestic violence rates. So part of that seems to be driven by maybe some household bargaining power dynamics. Anna Aizer has a paper from 2010 that looks at how reducing the gender wage gap also reduces violence against women. And then there are a couple of papers looking at male unemployment rates and female unemployment rates separately and how those relate to intimate partner violence and child maltreatment and find, you know, that there can be results consistent with that bargaining power story and also with a time use model where more time at home, especially for unemployed males, can be dangerous for their children. Then there's - kind of in another category - the Card and Dahl paper from 2011 that studies what happens when home football teams experience unexpected upset losses and they find these spikes in intimate partner violence around the time of the loss. So that shows us that unexpected emotional shocks can also be a trigger for domestic violence.

 

Jennifer [00:04:19] So what are the obstacles that researchers like yourselves face when you want to measure the effects of some event or intervention on the incidence of domestic violence?

 

Emily [00:04:28] The first things that come to mind for me are data related. So for us specifically and really for anyone who's trying to produce timely covid evidence right now, we need data sets that become available very quickly. So, for example, none of the data sets in the domestic violence literature that I just cited would work because none of those data sets are released yet for 2020. So that's definitely a limitation for the research teams that are working on covid related studies.

 

Emily [00:04:59] And then even outside of covid, measurement is tricky with domestic violence always. First, we might think of under reporting and under measurement, because, of course, we expect that there are domestic violence incidents happening that never make their way into any data set. You know, they're flying under the radar, unfortunately. And there's also the possible problem - the likely problem, of counting things as domestic violence that aren't actually domestic violence. And we should be cognizant of that in our own work because when we look at police calls for service, we do our best to identify likely domestic violence calls, but we don't have fine enough information to identify them perfectly. So those are both, you know, measurement and having quickly available data were two things that we had to think carefully about.

 

Jennifer [00:05:48] Yeah, I think a lot of people who don't work in this area would be surprised to hear that in 911 call data or reported crime data, there isn't just like a domestic violence category that is uniform across all cities.

 

Emily [00:06:00] Right. I - it's true. I mean that's something we had to go by - city by city and see what the incident descriptions look like. And some of them do have domestic violence or DV tags, but some of them are just domestic disturbance or family fight. And so if we actually knew exactly what those incidents looked like, we could parse them out or if we had more information on the identity of the caller or a lot of other pieces of information could be helpful. But we kind of have to make the best of what we have.

 

Jennifer [00:06:34] Okay, so let's go way back to earlier this year, which feels like a lifetime ago, I'll admit. Remind us about the sequence of events related to the pandemic and social distancing and stay at home orders. What is the timeline that you're considering in this paper?

 

Emily [00:06:48] The very first case of covid in the United States was confirmed on January 20th, and that was in Washington state. But in terms of when policy things really started to change, when behavior started to change, all of that didn't take off until the beginning of March. So if you look at the week of March 9th, in that week, the WHO declared the COVID-19 was a pandemic. Donald Trump declared a national emergency. The NBA suspended its season. And you start to see the dominoes falling in terms of school closures and students transitioning to online learning. So all of those headline grabbing events were happening. And then if you go into data where you can see how people were actually behaving in terms of staying home, going out to eat, you'll see big changes at the same time, around that week of March 9th, which is about the tenth week of the year.

 

Emily [00:07:40] So in our analysis, we're going to take that as the start of treatment. But we also do some checks to make sure that our results aren't too sensitive to the exact point that we chose. And for context, the first state level stay at home order didn't come until March 19th. So big changes came ahead of when all of those stay at home orders came online.

 

Jennifer [00:08:01] Yeah. So let's dig into that a little bit more. So you actually use a few different data sets in this paper to figure out when exactly people change their behavior. So what data do you use to do that and what did those data tell you?

 

Emily [00:08:13] We used four data sets. So we've got two data sets that are cell phone tracking data, one from safe graph, where we are observing the percentage of devices that don't leave home at all on a given day by census block, and then in unicast, that cell phone tracking data categorizes travel as essential or nonessential. So we can look at the amount of non-essential travel happening over time. Then we also have open table data that measures the number of seated diners in, you know, restaurants daily. And then Google Trends data tells us about search intensity for the term social distancing. And in every one of those data sets, we look at trends by states. So we make figures where you have a line for every single state and we can see what the patterns look like over time across states and across the different data sources. And what we see is a pretty coherent picture. It all points to increases in social distancing, decreases in leaving home starting again around the 10th week of the year around March 9th. And the timing is really quite tight across states. So we ran specifications where we assign treatment that was specific - the timing specific to each place, but it doesn't really make a difference. The pandemic effects are kind of breaking out everywhere at the same time.

 

Jennifer [00:09:33] That's really interesting.

 

Jennifer [00:09:35] Okay, so once you figured out when people's behavior changed, how did you use that information to measure the effect of social distancing on domestic violence?

 

Emily [00:09:44] Well, one thing that we needed to make sure we did was to account for seasonal trends. That was something that not all of the media reporting that I saw seem to be doing. But violent crime exhibits strong seasonal patterns. Specifically, there's more violent crime during warmer months. And that's true for domestic violence specifically as well. So we use data from 2019 to help us capture those seasonal trends, what we're doing is estimating a difference-in-differences model where we're comparing domestic violence calls for cities in our sample before and after that cut point of March 9th, the 10th week of the year in 2020y versus 2019. So it's a difference-in-differences where both differences are across time, you know 2019 versus 2020. And the first couple of months of the year, with the spring months of the year, we would expect to see an increase from winter to spring in domestic violence in any year because of those seasonal trends. So what we wanted to do was to assess whether there was an even larger increase in 2020, which we would then attribute to the pandemic. And the identifying assumption we need to make is that the trends in 2020 would have been similar to the trends in 2019 had there not been a pandemic. And so we use an event study in part so that we can check for parallel trends. So in the event city, we're comparing, you know, the first week of 2020 with the first week of 2019 week by week, the first 21 weeks of the year. We kind of plot out what the differences look like. And we see pretty clear trend break around that tenth week. And so the trends look similar leading up to that point. And after that there's definitely higher rates of domestic violence reports in 2020.

 

Jennifer [00:11:34] Okay, so what data are you using to measure the effects on domestic violence? You mentioned a big repository that you and Riley were looking at.

 

Emily [00:11:42] Right. So the police data initiative is an online repository that some police departments across the country choose to participate in. It basically has links to sites where you can go and download data, whatever data the police departments decide to share. And there are a couple of dozen cities that share police calls for service data. We had to focus in on cities that update their calls for service data frequently. So if the last update was in 2019, that wasn't going to work. We also needed to focus on cities that had enough incident information that we could identify likely domestic violence incidents. So those were the kind of constraints on which cities made it into our sample. And as I said before, not all incidents, of course, are going to get called in, and we're not able to perfectly identify domestic violence. And that's part of the reason why we wanted to focus on looking at percent changes, using specifications with logs instead of the levels.

 

Jennifer [00:12:41] Yeah. So why don't you tell us a little bit more about how you actually figured out - or to find what is a domestic violence call? How did you go about that?

 

Emily [00:12:49] So in a given city, we would find the variables - I don't know one or maybe two variables that would have the incident description. And then, you know, it was kind of a laborious process that first of me tabulating these incident descriptions and just reading them to see what they looked like. And as I did that, I started to see patterns of the word domestic, the phrase domestic disturbance, or DV. So then we're using regular expressions in data to kind of tag those incidents that have one of these key phrases in the incident description. So, you know, it's not something that you can automate from the beginning. You definitely - the messiness, as you know, of administrative data requires a certain amount of doing things by hand at first.

 

Jennifer [00:13:37] Yeah. And again, for those who haven't looked at these data themselves. When we're talking about the different crime categories in those - in, you know, the description, it's not like there are 20. Right. There are usually like hundreds, and there are like misspellings - spelled with one 's', spelled with two 's'.

 

Emily [00:13:54] Yeah. You got to worry about lowercase, uppercase, spaces on the beginning and ends. So there's - yeah, there's a lot of little, little ways that you can miss what you're trying to catch.

 

Jennifer [00:14:03] Yeah. It does make you wonder if like - if all these departments just use like a drop down menu or something, it would make their lives so much easier. Instead it's clearly someone just like typing in like what the incident is.

 

Emily [00:14:13] Yeah. If anyone wants my input, I've got some suggestions on ways to make it more researcher friendly.

 

Jennifer [00:14:20] Love it. So which cities wound up being in your final sample and I guess bigger picture, are these cities unusual in any way relative to other urban areas in the US?

 

Emily [00:14:28] So the cities that we ended up with were Baltimore, Chandler, Cincinnati, Detroit, Los Angeles, Mesa, Montgomery County, which is the only non-city entity, New Orleans, Phoenix, Sacramento, Salt Lake City, Seattle, Tucson, and Virginia Beach. And all of these cities are in counties that - at least near the beginning of the covid outbreak in the US, had above median cases per person, and six of them were in the top quartile. So they may be places that were a little more intensively treated in the sense that the pandemic was more - that there were more cases of covid in the cities. There are also - you know, we have a variety of geographic areas and regions in the data, but it's not perfectly representative. There are several cities from Arizona, you know, only two from California. So there's variety, but I wouldn't call it totally representative sample.

 

Jennifer [00:15:28] So let's talk about the results. What do you find are the effects of social distancing on domestic violence?

 

Emily [00:15:35] We found that reports of domestic violence were 7.5% higher from March to the end of May. And the increase was concentrated during the first five weeks when the effect was closer to 9.7%. We also found that if we had failed to account for seasonal trends, so if we just hadn't used 2019 at all as a control year and just kind of compared early 2020 with March through May 2020, we would overestimate the effects by about 100%. So it's an important adjustment to make. But even accounting for seasonal trends, you know, 7.5% or 10% increase in domestic violence is still significant and large enough, I think, to be concerning.

 

Jennifer [00:16:16] And so these effects kind of change over time. Right. So describe what you see in that event study graph where you're kind of plotting it by time.

 

Emily [00:16:23] Yeah. So if we look at just where the point estimates are falling in the event study, there is a bit of a drop off around the 15th week of the year, and then see the estimates rise again after that. You know, we don't have enough precision to say for sure that this is a meaningful drop down during that 15th week. But if we wanted to think about possibilities for why might there have been a drop around then, there's at least two things that we could think of. One is that when we look at the safe graph data that tells us what share of mobile devices we're staying completely at home, we see a drop off in social distancing around the same time. It's not huge, but it's possible that households that are at risk for domestic violence would stop social distancing relatively sooner than others. So that could kind of explain the pattern that we see.

 

Emily [00:17:14] Another thing that happens that week is that most of the CARES Act relief checks went out on April 15th, right in the middle of that 15th week. We don't have data on who is getting checks in each part of the country, but Raj Chetty has data that shows that that day, April 15th, was when the vast majority of those checks went out. So financial relief from the stress around job loss may have played a role in helping to mitigate some of the domestic violence effects.

 

Jennifer [00:17:46] Yeah, which is really important when we think about - well, I'm sure we'll get back to this when we talk about policy implications - of all of this, and what we could be doing differently.

 

Jennifer [00:17:54] So you run a bunch of additional tests to make sure that you're isolating the effect of social distancing from other factors. So tell us about a couple of those and how you and Riley are able to convince yourselves that the effects you're measuring represent causal effects?

 

Emily [00:18:08] Yeah, this paper was very focused on establishing mostly one empirical fact, which is how did the pandemic effect domestic violence? So we kind of wanted to throw everything we could at it, make sure that we were - that the effect we're measuring was really robust. And - and it is. One of the things that we did, in addition to some of the checks I mentioned earlier, was to re-estimate the results by excluding each city in the sample one by one. So - like, Los Angeles is huge. We want to make sure it's not just one big city or maybe a city with a really extreme pattern driving the estimates. But when we exclude each city, the coefficients are really stable. So that's confidence building.

 

Emily [00:18:50] Another thing that we did - my personal favorite was some placebo tests. So for the cities where we have 2018 data, which excludes Detroit, we picked 100 placebo treatment dates during 2019, and then we reran our model so that we're comparing 2018 versus 2019 before and after the placebo treatment date. And we plotted the coefficients that we got from those 100 placebo regressions, and also plotted our main result on a histogram. And you can see in the figure that our estimate is a clear outlier. And it's even more of an outlier if we re-estimate our covid effects using 2018 as the control instead of 2019 to make sure that the samples are comparable. So what that tells us is that the pattern that we see in 2020 is definitely outside the range of what we would expect from just normal year to year variation. And another way of thinking about it is that it looks like 2019 is a pretty reasonable control year because the differences between 2018 and 2019 are relatively small.

 

Jennifer [00:19:53] So one of the toughest aspects of studying domestic violence, as you mentioned earlier, is that it often goes unreported. And so we always worry that changes in an outcome like 911 calls might represent a change in reporting rather than a change in the actual incidence of the crime. So in this case, changes in reporting might bias your results upward or downward to make it extra complicated. So tell us about that. What should we be worried about here?

 

Emily [00:20:18] Yeah. Absolutely. Reporting is a major component of the measurement issues that I mentioned earlier. I think in general, we should expect domestic violence to be underreported. If you look at the National Crime Victimization Survey, which may itself suffer from underreporting, about half of the intimate partner violence incidents that show up in the survey were reported to police. So that just gives you a sense of the magnitude of underreporting we're dealing with. In our study, something that we need to keep in mind is the possibility that the treatment could have had a direct effect on reporting behavior and that could lead to a bias even in our percent change results. So if victims became less likely to report during covid because say it's harder to find the privacy to do it, then that would lead us to underestimate the true effect of covid on incidents.

 

Jennifer [00:21:08] So basically the story you have in mind here is that if you're the victim, your partner's home all the time now, and so you never are able to call the police because they're always in the same room. Is that basically it.

 

Emily [00:21:17] Right. Yeah, something like that. I mean, and it's also, you know, if you - victims may need support from people outside of their household to seek help that they need. And if they don't have access to their support system that could play a role as well. Yeah, I think greater isolation for victims is probably never a good thing as far as helping them to get support and escape from the situation they're in. On the other hand, if third parties like neighbors are more likely to report during covid because, you know, if everyone's home, then if you - you know, you're around to hear what's going on next door. And that could lead us to overestimate the true effect of the pandemic if there was enough additional third party reporting because of shelter in place.

 

Jennifer [00:22:08] So you do something very clever to test for whether a rise in third party reporting, so that's calls from neighbors, say, whether that might explain the rise in domestic violence calls that you're seeing, and you use the type of housing that people live in. So tell us what you do here and what you find.

 

Emily [00:22:23] Yeah, we had to get creative because as I said earlier, there's - we don't have as much detail in the data as we would like. We wish that we could just tell from the call whether or not it came from a third party, but we can't. So what we decided to do was to look at the neighborhoods. in most of the cities, we can observe something about the location, not the exact location, but something that allows us to link to census tract like what city block it occurred on. And then we compare census tracts that have a lot of multiunit housing compared with neighborhoods with very little multiunit housing. And our logic there is multiunit housing means shared walls. That means it's easier for neighbors to hear what's happening next door versus in neighborhoods that have a lot of detached homes. So if there was a big third party reporting effect, then we would have thought there would be larger effects in those high multiunit housing areas. But in fact, our point estimates for the two groups are nearly identical. And that suggests that third party reporting probably isn't driving the effects that we're estimating. It also lines up with some statistics that the National Domestic Violence Hotline has published. They have seen an increase in call volume during 2020 compared to 2019, but the fraction of calls to their hotlines from third parties has been very similar between the two years.

 

Jennifer [00:23:45] So you also consider whether effects vary across different types of neighborhoods to shed some light on who is bearing most of the burden of this increase in domestic violence. What do you find there?

 

Emily [00:23:55] When we look across different racial or socioeconomic status groups, the biggest takeaway for me is that we're seeing evidence of effects almost everywhere. We are losing precision on some of those groups because of the smaller sample sizes. So the differences aren't big and we can't rule out that the differences we do observe are just statistical noise. I was surprised to see that if we look at neighborhoods with high versus low changes and time spent at home or high versus low predicted layoffs, we don't get a clear picture of some areas being much harder hit than others. And my best reading of those results is that the effects of the pandemic were just really pervasive. That's not to say that the costs of the pandemic are being shared equally by everyone, but effects are widespread. Every neighborhood is being hit by something, and it seems to be translating to higher risks for domestic violence pretty much everywhere.

 

Emily [00:24:52] Where we do see some meaningful patterns are first in the timing. So we see larger effects on weekdays, which is logical if we think about school closures and working from home - times when people aren't used to being together. And also when we tried to get at the extensive versus intensive margin, so if we look at city blocks that have a recent history of domestic violence calls versus those that don't within the past, say, three, six, or 12 months, we see large significant increases from blocks that did not have a recent history of calls. So that looks like new households, an extensive margin increase versus the - repeat blocks where we have negative but pretty noisy estimates. So we can't rule out that they also experienced large effects. But we can be really confident that there was an extensive margin effect - that new households were getting involved in domestic violence.

 

Jennifer [00:25:53] So this has become a very hot topic in recent months. There are lots of people working on papers related to covid and domestic violence. I'm not sure how many there are or how many are in working paper form at this point. I've lost count. What other studies do you know about and how have those results compared with your own?

 

Emily [00:26:11] Yeah, there's lots of great contemporaneous work. And from the papers that I've seen, and I'm sure I'm not aware of all of them either, but the ones that I've seen so far have findings that are pretty consistent with ours. There are a couple of papers that look like at a single large city, at crime in general, at different crime types. Those don't all find statistically significant effects, but we have more power to detect smaller magnitude effects because we're pooling across cities. And the papers that do have multicity data like ours find similar results.

 

Emily [00:26:45] So to highlight a few of the ones I know about, there is a paper by Bullinger, Carr, and Packham that uses data from Chicago specifically, and they have the police calls for service data, but they're also matching it up with reports made by the police - the incident reports that were filed by police. Not every call leads to an incident report. And they see the increase in calls like we do, but then a big decrease in reports from the police. And there are several possible reasons for that and that they discussed. But I think it's really helpful to think about the different layers of reporting and what the patterns that they see could tell us about what's going on. There's also a paper by Sanga and McCrary that uses again calls for service data. A lot of overlap with our data set. So they - they see the same increase that we do, and then they do a really nice job of developing a framework for inferring information about household extensive-intensive margin effects from neighborhood level data. And using that they find the same thing that we did where there's larger increases on the extensive margin - so new households versus the intensive margin, repeat offender households.

 

Emily [00:27:58] There are a few papers - I'm excited to see more in this category that tell us something about specific policies and how those play into the domestic violence effects. So there are a couple of papers from India where there was geographic variation in the severity of the lockdown. And there we see like in the paper by Ravindran and Shah that there were increases in domestic violence complaints in areas with the strictest lockdown. So in our context, we really couldn't disentangle all of those the lockdown from the school closures and everything else because it was all just layering on top of each other and there wasn't variation across places. So I think it's great to have evidence in a context where you can do that. And, you know, there's - the last one I'll mention is a paper using Italian data by Colagrossi, Deiana, Geraci, and Giua that finds that an anti abuse campaign during covid in Italy increased hotline calls. I think that's really helpful to realize that there are ways of reaching out to victims and trying to induce more of them to get support for what they need. This is a specific policy that's probably usable in other areas as well.

 

Jennifer [00:29:10] I realize that I should have mentioned upfront that your paper is forthcoming at the Journal of Public Economics, if it's not already published there. I imagine we're going to need a couple of separate volumes to fill - to take up all these other papers. So hopefully some enterprising journalist.

 

Emily [00:29:25] That's right. But it's kind of great from like a social perspective. It's - as a researcher, it's stressful to be competing with other researchers. But from a policy perspective, it's great to have multiple takes on the same question because then you can really - especially when they're so consistent, you really know what's going on.

 

Jennifer [00:29:43] Absolutely. And so speaking of that, what do you think policymakers should be taking away both from your study and the other work in this area? What are the most important policy implications here?

 

Emily [00:29:53] So from our study, I would say, first of all, that stay at home orders are not going to be an on-off switch for these domestic violence effects. When we look at our urban studies, so we can see week by week what the effects looked like. The increase happens ahead of the official stay at home waters, it coincides with when behavior is changing. So it's important to realize that we can't just control the domestic violence risk using those shelter in place policies entirely. It's also, I think, important to know that we can't only focus on known or previously reporting victims. We have to take into account that there are new potential victims out there that are also at risk in these situations.

 

Emily [00:30:38] I would love to know whether financial relief like the CARES Act checks was actually a factor in reducing some of the domestic violence effects temporarily. It's consistent with our results. I don't think we can say we've definitively answered that question, but I would love for policymakers to implement policies in that direction and perhaps do it in a way that would allow us to figure out if it's helping or not.

 

Jennifer [00:31:05] Yes, some staggered timing for some of these checks would be perfect. And have we learned anything that can inform our approach to domestic violence during normal, non-covid times?

 

Emily [00:31:15] Well, you know, I'd say that - 

 

Jennifer [00:31:18] I'm being optimistic and assuming will eventually get back to normal, non-covid times.

 

Emily [00:31:24] Yeah. That's right. But I mean, I think that we are learning that - I guess it's adding, in my view, to some of what we have in literature already that labor market dynamics matter a lot. You know, emotional stress matters a lot. If it was just about time at home, then when we looked, we could see differences based on which neighborhoods are spending more time at home, and we just don't see them that clearly. So I think all of the strands in this tangley knot, I expect, are having an effect on domestic violence rates. And I think that speaks to the complexity of the problem that there's not just one thing to approach, to try to fix it, but maybe also gives us hope that there are multiple angles that we could come at to try to mitigate - you know, improve conditions for people who are at risk.

 

Jennifer [00:32:15] Yeah, that's a great point. And for your earlier comment, we've also learned we need better data.

 

Emily [00:32:20] Oh, yes.

 

Jennifer [00:32:21] We need - we need dispatchers and 9 - CAD systems to include better data about which calls are domestic violence calls.

 

Emily [00:32:28] Yes.

 

Jennifer [00:32:30] So what's the research frontier here? What are the next big questions in this area that you and others are going to be thinking about in the years ahead?

 

Emily [00:32:36] Well, I would love to see research on what policies work to counteract the increase in domestic violence that we see in this specific covid situation, as well as in other stressful or economically difficult situations that households are in, whether financial relief is effective in helping there or not. I would also, as far as covid goes, love to know more that helps us to isolate the different strands of the mechanisms - you know, the school closures, working from home, being out of work. Are all of these things contributing equally? Are there some groups that are more affected by them than others? So I think that there's a lot - you know, getting under the hood of this reduced form stuff that we're seeing and also knowing what we could do to fix it - trying things to fix the domestic violence increase, and then seeing what works would be top priorities for me.

 

Jennifer [00:33:31] My guest today has been Emily Leslie from Brigham Young University. Emily, thanks so much for talking with me.

 

Emily [00:33:36] Thanks for having me.

 

Jennifer [00:33:43] 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. 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 Jonn 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.