Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Learning to Harm: The Intergenerational Transmission of Gender-Based Violence

Authors: Bilge Erten, Susan Niknami, Emily E. Nix, Anna Sandberg  
Discussant: None  
Video: https://www.youtube.com/watch?v=1kb3a99sA-E&t=3390s


## Talk (00:56:30 – 01:58:19)

[00:56:30] and in the uh universities that we work in. Thank you. [applause] >> Uh so Emily  
[00:56:56] Everything okay? >> Perfect. All right. Well, so first I just want to say thank you to the organizers for putting us on this amazing program. I've already learned so much. What a fabulous paper just before this one. Um, this is joint work with  
[00:57:11] Billy Erin who's in the room and is going to help keep track of all your brilliant questions. Susan Nami and Anna who are in uh Sweden and Sean who just moved to Pit and he's an amazing uh sociologist working at the Census  
[00:57:22] Bureau. Um, so the big picture of this question is we know that I don't have to convince this room that violence against women is a very prevalent crime. So if you look at survey data, one in three women are impacted. This is probably an underestimate if you look at some other  
[00:57:37] surveys. So a lot of women experience violence over the course of their lifetime. Now it's also true that these uh crimes are very costly. So I spent uh the greater part of the first part of my career trying to explore to the degree to which this is not just a moral crime  
[00:57:52] but it's actually an economics crime that there are real economics consequences that there are strategic economic impulses at play when it comes to domestic violence and so on. Now the problem though is once you have a crime occur, you can try and do everything you  
[00:58:06] can to help a victim. But in an ideal world, we would not have these crimes occur at all. And so what I'm hoping to do in the the next part of the career is think very carefully about how do you stop violence against women from happening in the first place. Um because  
[00:58:20] that's the best way to reduce these economic costs. We know that these have dramatic consequences on women's careers. We just saw this in terms of major choices. Um I've document a lot of the uh employment consequences of of rape and sexual assault of domestic violence of workplace harassment and so  
[00:58:34] on. But ideally we can stop those economic harms from happening in the first place and that could have real implications for GDP for outcomes for women across the world. And then this leads to a central open question, which is, is violence against women something that is unsolvable that we're just going  
[00:58:49] to have to deal with events when they happen and try and do our mo our utmost for victims or can we have a deeper and better understanding of where this behavior comes from? And with that understanding, can we then come up with policy levers to pull to reduce these  
[00:59:03] crimes in the first place? And given how prevalent they are and how costly they are, I think that's the major question.  
[00:59:09] And so we're going to try and answer two main questions. uh in this paper and and before I get into these questions to motivate them I'd like to give you an anecdotal account that I think really uh brings home why we're so interested in this question. Um so increasingly I've been trying to work a lot with policy  
[00:59:24] makers, cops, judges, prosecutors and so on to kind of understand what they're struggling with and bring the amazing research that's happening from folks in this room to those people so they can use them in their everyday work. Um, and I was speaking with an amazing woman who's a cop in Texas and she runs a  
[00:59:38] large domestic violence unit and I was we were talking about the very early seeds of this paper. Billy and I were at a conference with this amazing cop. Um, and what she said as soon as we mentioned this idea is, "I just actually had a stop. I arrived at a home. A man had been brutally beating his wife. I  
[00:59:52] arrived at the house. He's uh yelling at his wife that she has to get his shoes.  
[00:59:56] I'm trying to diffuse the situation and say, "No, you don't have to get your shoes. Uh, his shoes. I'm taking him now." and his young teenage son walks up, slaps his mom, and says, "You have to get her shoes." This is a horrible story. But what she told me is, I, you know, I then interacted with the kid who  
[01:00:10] I now have to arrest and bring to jail because I've witnessed an assault. And I don't think at his heart this is a bad kid. I think this is a kid who has seen his father show demonstrate this behavior over 10 12 years and is now mimicking the behavior himself. All  
[01:00:24] right? And so what we're going to do in this paper is we're going to first in the first part of the paper do a descriptive exercise of how strong is this correlation. How often do you see sons who have a father who has ever committed violence against women go on to commit violence themselves? How often  
[01:00:38] do daughters who are exposed to a father who has ever committed violence against women go on to have partnership outcomes that put them at greater risk? That's going to be a pure descriptive exercise.  
[01:00:47] So think of interracial transmission of income but instead looking at violence against women. That's going to I'm going to go through that fairly quickly because then we're going to jump into the uh in some ways more interesting part of the paper which asks can we undo this transmission. All right. If you is  
[01:01:01] this transmission caused by the exposure to the violent father and if so does removing him decrease the transmission and that brings us to potentially hopefully good policy levers we might be able to pull in in these cases controversial but policy levers we can  
[01:01:16] pull. So we contribute to three main literatures. So number one is an enormous literature on early child environment and outcomes. So think Janet Curry and Anna Eiser's incredible work um and we'll see more of this later in the week in the children's program and they've done amazing work along with  
[01:01:31] many other people showing that the environment and what you grow up has huge implications for how you behave, how you interact with the world and your economic outcomes later in life. And so we're going to study a very distinct relevant margin which is how much are you exposed to a male role model who has  
[01:01:45] ever exhibited violence against women and how does that change your perpetration and victimization rate risk into adulthood. Um second we're contributing to a smaller but growing literature specifically looking at violence against women. Lots of more papers looking at costs nowadays but a  
[01:02:00] few papers trying to understand how do you reduce it. Sizer has some great work here um about economic outside options for example. There's also several wonderful papers. Sonia has some work on this um looking at the role of police.  
[01:02:12] Um and so I think those are all really important. We're going to try and go back to the very earliest stages of where does this behavior even come from amongst children.  
[01:02:21] Last there's an amazing fascinating uh set of papers looking at the intergener generational transmission of crime. A great paper that I think is underststudied, I believe it's in labor economics by Randy Yammerson and Matt Linkfist in Sweden looks at transmission of other types of crimes like property  
[01:02:36] crime, assault, high transmissibility. Um, a set of other papers looks at what happens when you remove a dad to go to prison for other things. What happens to child outcomes? Um Jeff uh Jeff Weaver, my co-au at USC and co-authors find a  
[01:02:49] very famous null. Um Suzanne in a Suzanne in her paper in Sweden looking at incarcerations for other reasons finds that it's quite harmful normally to remove a dad on child outcomes. And we're going to see if that holds true when you have this particular type of  
[01:03:03] death. All right. So um to talk about data. So I had promised the organizers we would have data from the US. I'm sorry I lied. Um for various reasons related to disclosure changes and so on.  
[01:03:15] That is still in progress. It's going to be there eventually. Um so today we will focus on the Swedish data. Um I don't think I 100% guaranteed it in my in my uh when I filled in the form and submitted the paper. Um [laughter]  
[01:03:27] so you know academics um but we will I'll talk very briefly. We do have uh we will eventually have some data on the US. Um we're going to be able to do everything we believe but the judge fixed effects. Um but I can't say anything more on that at this point but keep stay tuned for that. Hopefully  
[01:03:41] sooner rather than later. So I'll focus on the Swedish data. This covers individuals from 1972 to 2022. Um, that big window is really necessary when you're trying to look at intergenerational outcomes for crime.  
[01:03:53] Um, we have their normal amazing Nordic data where we have all your demographics, household composition, etc., etc. And we link that to the Swedish crime suspicions register and the conviction register from 1973 onwards. And this provides us some  
[01:04:06] information on criminal involvement. Now, let me define our main measure, which is a father ever suspected of violence against women. In the Swedish data, we do not know the victim of the crime. Instead, what we use is a very  
[01:04:20] helpful set of crime codes that says it was assault of a woman. It was violation of a restraining order, which is often a restraining order that a woman has put out against a man, sexual violence, including rape and molestation, and gross violation of a woman's integrity.  
[01:04:34] As I'll show you in the next slide, the vast majority of these men will be accused of assault of a woman. Um, our sample is going to include over three million father-son and father-daughter pairs. And as our main outcomes of interest, we are going to ask if you are  
[01:04:49] exposed to a father who ever commits one of these crimes. And in robust checks, don't worry. We also look at if he commits it when the child is under 18.  
[01:04:56] But for for a proxy, we're going to say, do you ever commit one of these crimes?  
[01:05:00] um do we see that the son is more likely to commit to be suspected of committing violence against women himself using these same crime codes? And for daughters, we're going to primarily use a victim victimization risk measure,  
[01:05:14] which is does the daughter cohabit with a man who is ever suspected of one of these crimes. Now I want to pause and say we're we're using the ever suspected based on some prior work of mine which suggests that there isn't a type that this you know assigning a men who ever  
[01:05:28] commit violence against a woman is a type seems accurate. Again we will have robustness checks that go from 0 to 18 to just clarify if it happens when the child is a child.  
[01:05:37] >> Y >> I know this is descriptive but it still might be interesting to compare to like a similarly large crime category. So like property crime and because it might be just that like >> I have an opinion table for you. Yes. Um uh so the first part of the paper is really just laying laying out these  
[01:05:51] descriptive results about intergenerational transmission of BAW and we have an appendix table where we say okay let's take other crimes using Randy and Matthew's paper let's look at income transmission look at look at transmission of all these other things including other major crimes  
[01:06:05] >> transmission is causal language so versus just like correlation of like socioeconomic status that leads to similar right and so I think that's where comparison to like other types of criminality would be really helpful >> yeah and I can if I have time I'll show the table where we do that. Um what I  
[01:06:19] would say the reason I'm saying transmission is the first part of the paper purely descriptive. Second part of the paper I'm going to convince you I hope that is causal and so we're going to be able to say this actually does seem to be transmission from exposure to the father and I'll show you a lot of evidence that I think very clearly  
[01:06:33] points to this being actually causal. And that's where we're able to go beyond say you know a big literature that just focus on how transmissible is income.  
[01:06:40] We're going to say violence against women very transmissible and caused by exposure to father. Yep.  
[01:06:49] So I think there is a very interesting question about uh nature versus nurture here. And so uh you create the um exposure to victimiza victimization index for daughters but can you see um  
[01:07:01] for for sons can you see whether they uh live in a family where the partner of a mother who's not their father is >> absolutely. You're getting way ahead of me. Wait till like slide 30 and I I'll get to that point. But yes, we're I  
[01:07:16] think we're going to talk about several mechanisms by which this transmission could occur. Um, some of which are going to be caused by exposure to father if we see it descriptively in the first part of the paper, some of which would not.  
[01:07:25] And then we're going to spend the second part of the paper digging into that. So, brilliant idea. We agreed. We'll show you results on it. Um, so we'll show you a lot of results on other male role models in the family, etc., etc. All right. So, like I said, most of these men are committing crimes of assault  
[01:07:38] against a woman. Um, similarly with their sons who go on to commit uh violence. This is conditional on committing violence. What is the violence category? Just so you get a sense of it. Now, I'm going to jump straight into these descriptive correlations. All right. So, on the blue  
[01:07:52] bar, this is sons whose fathers are never showing up in the police reports as suspected against violence against women. Even among those sons, almost 6% show up later with the police reports of suspicion of violence against women.  
[01:08:05] Now, that being said, if you are a son who grew up in a household where your father was ever suspected of VAW, you are three times more likely to go on to be suspected of violence against women yourself, that's a really high, just baseline raw share of sons who go on to  
[01:08:20] commit violence against women. And this is just raw data. We can throw a bunch of cor, you know, controls in there, do a match propensity score matching exercise. If you go to that baseline of five about 6%, we s find in the raw data you're 11 percentage points more likely  
[01:08:33] in LPM model to commit violence against women. If you're a son who grows up in these families, even if you throw everything in the amazing Nordic data at that, it only drops down to just under 8 percentage points more likely. Now, again, this is all descriptive. We're not saying anything's causy yet, but these seem to be very, very persistent  
[01:08:48] and strong correlations and in exposure of sons to perpetration of violence.  
[01:08:52] Yeah. >> Are you able to use radiation on old.  
[01:08:59] >> We're going to do that in the very last slide. Uh second to last slide, you guys, I love this room. You have great ideas. All right. So for daughters, when we think about and and you know, I'll quote Jana. If you don't like LPMs, we do odds ratios. It's all fine. Um all  
[01:09:14] right. So if we look at daughters and victimization risk, our main measure is going to be do you cohabit with a man who is ever suspected of violence against women? And we see that among daughters whose fathers never are suspected against violence against women, about 8% of them end up  
[01:09:28] cohabiting with a man who is suspected of VW. That's twice as large, about 16% for daughters who grew up in homes where their father was ever suspected of VW.  
[01:09:37] It turns out these daughters who grew up in households with violent fathers are also more likely to partner in general.  
[01:09:43] So we can also do the conditional raw means and we see that then it's maybe even larger. Um now I will also say that one big concern we have which Sonia already mentioned in the last paper is reporting. Uh these are very unreported  
[01:09:57] crimes. So one thing we can do to cross validate and we're most concerned about that by the way with these first uh part of the paper with the descriptive correlations to cross validate these measures. We've gone to the health data and we can look at flags for domestic  
[01:10:09] viol for IPV. Now I will warn you that you know this is a very low rate as you see on the yaxis. This is because you have to have been assaulted so horrifically that you show up in the health system and they also have to be confident enough that IPV to code it as  
[01:10:22] such. That being said, if anything, it seems like these this correlation, this increase in uh risk for daughters who grew up in violent homes where fathers were suspected of violence against women. If anything, we might be underestimating it with the police data  
[01:10:36] because we find that the likelihood you show up with a hospital or healthc care center flag for IPV is four four times higher for daughters who are exposed to a father who ever is suspected of violence against women. Yep.  
[01:10:48] >> Is that good or bad? >> So, we haven't done we're looking on we we do that in the the judge IV which I'm going to get to in a little bit. We remove the child abuse cases. Um it could be that not only is he beating the his partner but his daughter and we see  
[01:11:02] pretty frequent child abuse cases amongst these men as well. And so we could we haven't done that. That's a good idea. We can do that for this data as well.  
[01:11:13] >> Oh, right. Yeah, that's right. We can't see the victim flag here.  
[01:11:17] >> It's not. >> Yeah. Yeah. Yeah. Sorry. That's right.  
[01:11:21] The it's the health center. So we can we'll dig into it and see if there's anything else we can find in the flags and the health data. Again, if we throw everything into the model and LPM model, it doesn't budge that much. It does go down. You know, some of this is correlated with other things and it decreases, but you're still six  
[01:11:36] percentage points more likely to partner with someone who has ever been suspected of VW if you grew up in a home where your father was ever suspected of VAW.  
[01:11:45] Really quickly before I go to the meat of the paper to the more causal part of the paper, we were also very interested in partnership outcomes in general. Um since especially for daughters, these results are manifesting in in large part through a partner choice. We were  
[01:11:58] curious about do we see for example assortative matching and we do. So if you look at that middle column right here, you can see that for both sons and for daughters, they are significantly more likely to partner with someone  
[01:12:12] whose father was also suspected of violence against women. So we see an assortative matching on this norm, whatever this norm is, that's being transmitted, it seems, across generations, at least in the raw descriptive data. We also see that both  
[01:12:25] sons and daughters are actually more likely to show up with any partner in the data. um which you know there's lots of reasons we can come up with daughters. This is a bit surprising to me at least for sons that they're one percentage it's not huge but they're one percentage more likely to show up with  
[01:12:38] any partner and last their partners are more likely to have less than high school. Um we think this is partially a descriptive fact of the of the of the idea that if you look in data on domestic violence violence against women consistently you see that this is negatively selected on observables.  
[01:12:52] Again, this might be a reporting thing as opposed to an actual incident effect, but you know, it's just useful to to see what the partnerships look like.  
[01:13:00] >> Did you do the match here? >> We have uh this is the matched one. Oh, no. No, this is not the matched one.  
[01:13:06] >> This is the matched one. Yes. Thank you, V. Yes.  
[01:13:10] >> Yeah, we we've gone back and forth about which one we put in the main text, but the everything looks similar with the raw data as you could see with the LPMS.  
[01:13:17] Yeah. All right. So first 15 minutes hopefully what I've shown you is that there is an extremely strong transmission. I don't have time to do the table but it's as strong if not stronger than some other things we really care about that previous  
[01:13:31] literature has documented in terms of say trans it's not stronger than transmission of education but it is about as strong as transmission of some of these other crimes. Uh, but we do think this is a a unique characteristic that could be transmitted. And what we're going to be able to spend the rest  
[01:13:44] of the paper focusing on is the extent to which this is caused by the exposure to the father who committed violence against women and the extent to which you can break this transmission uh through reducing.  
[01:13:54] >> Can I just ask a question? If I take the Swedish population let's say in uh in the middle of your period which is what about 1990 or something?  
[01:14:05] >> 70. Yes. 80s late 80s. and I take a I take a 50-year-old man and I what is the probability in your entire life that you are ever suspected of IPV? What would it be?  
[01:14:20] >> So I haven't done that but looking it's going to be some combination well you could think of similar to some com weighted combination of these two. Um, so but I we can do it. We can give you the exact number. But if you look here, sons who grow up in families that never have VAW,  
[01:14:34] >> it seems extraordinarily high for just suspected.  
[01:14:42] >> Yes. >> Okay. Don't you think it seems high?  
[01:14:45] >> So this is the police admin data linked to the demographic data. So it's what the data says. So what I this is suspicions, not convictions though.  
[01:14:52] >> It's their role. >> Yeah. So any kind of violence where the gender of the victim is a woman is here.  
[01:15:00] >> Assault of a woman. >> I guess I guess yet another example of where I live in a bubble.  
[01:15:07] >> Okay. >> Yeah. And what I would tell you Claudia Claudia what I would tell you is that if you look at the data having you done a lot of work on this in Finland for example, it is we we do live in a bubble because suspect suspicions of violence against women is very common amongst  
[01:15:21] lower income families. Very very common. Right.  
[01:15:24] >> Um and so what I will say >> there's also alcohol and so for this period in alcohol in Finland was enormously high.  
[01:15:31] >> Yeah. And Suzanne and Aritzo and uh forget the other co-authors have a great paper in Sweden looking at the relationship between alcohol and IPV.  
[01:15:38] Very strong relationship there. Really important. Um not something we're looking at specifically here. Um the the relationship with alcohol and BW, but that's that is definitely a major cause.  
[01:15:49] But I would say we live in a bubble. I I think most of us know or have experienced violence against women ourselves. That being said, if you look at the observational data, um lowincome women are both especially vulnerable and show up much more in reports. Now,  
[01:16:02] again, we don't we don't know underlying incident. It could be that high-end women are less likely to report for all sorts of, you know, social reasons. Uh but yes, it's it's higher than I expected, especially so what I would say is when we first got this number here,  
[01:16:16] this is devastating to me that this is so high even among sons who are never exposed to reported VAW. Um so I I entirely agree with you Claudia. I would say this is what the data says. In US we will only be able to look at convictions, not suspicions. So it's  
[01:16:30] going to be a smaller base smaller base, but I don't that actually I think is speaking to the fact that we're missing a lot in the US, which is why it's nice to have both countries in this paper.  
[01:16:38] Um, so we can see both the very beginning of a of a police investigation and the end. Um, all right. So, what we can take away so far, sons of uh, fathers who are ever suspected of EAW  
[01:16:52] are much much more likely to become perpetrators, although the baseline is also still high. Um, daughters are much more likely to go have it with risky partners and much more likely to experience actual IPV and health reports. But going back to Karin's point, these might not be caused by  
[01:17:07] father exposure. It could be caused by all sorts of other things that we can't move with policy. And that would be a very depressing result. So we're going to spend the rest of our time is trying to understand a to the extent to which this is causal and specifically if you remove a father a father who is  
[01:17:21] suspected of EAW, do you see a reduction in some perpetration and daughter victimization risk? All right. Now there are several mechanisms that would say yes that should happen. So you can think about when you remove the violent father, all sorts of norms in the household change. So you know, maybe he  
[01:17:36] was speaking derogatorily about women all the time, beating your mom in front of you. There are all sorts of social learning mechanisms that can happen in the home, just like this cop from Houston was telling me about. At the same time, you can imagine that when you  
[01:17:48] remove a a horrible partner that the mom is able to recover. She's able to have a better life and she's able to provide better care for her kids. So these are kind of a bundle treatment. Whenever you remove a dad, you'll remove both the social learning and potentially help the mother have a better uh quality of life  
[01:18:03] and provide better parenting. At the same time, if you remove a father specifically for VAW, you could have a deterrent effect. So, it could translate to the daughter, this is unacceptable behavior. I will not accept this in a partnership. From the son, you could think, oh, I've seen that there's  
[01:18:17] repercussions for this behavior from the state. Therefore, I will not perpetrate it myself. On the other hand, there's lots of explanations that would tell us that removing fathers won't help. What if all of this is caused by neighborhood? What if all if it's caused by a broad set of social norms amongst  
[01:18:32] your community? Um, what if all of this is caused by some unobservable characteristic or observable characteristics, although we control for those that we can't control for in our LPMs? On the other hand, someone mentioned the this innateness of it may what if it's just there's innate tendency towards violence which can't be  
[01:18:46] undone through removal of uh role models. And last, when you remove a dad, many other papers, including a great paper presented last summer on divorce, find that removing dads is really bad for child outcomes. So even if this  
[01:19:00] particular facet is negative, learning to be violent, maybe the rest of the package of removing a dad, lower income, for example, is so much worse, that outweighs any benefits from the bad male role model being removed. And just to put some numbers on that, we replicated  
[01:19:14] that uh great event study from that paper that's I think RNRG right now on divorce. And and no surprise when you move a dad in these families, these violence against women families, you also see a big drop in household income.  
[01:19:25] And there's lots of papers on domestic violence, including some by those in this room, that shows that income shocks cause stress. Stress can cause conflict amongst partners, maybe the future partner of the mom, and that can itself actually increase domestic violence in the family, and maybe increase  
[01:19:39] transmission. Yep.  
[01:19:44] >> Dennis, >> what is what is the reason that >> Okay. What is the reason that you're specifically calling it like uh VAW fathers givingven the fact that if I'm a child, it's more likely for me to be  
[01:19:59] exposed to my father beating up my mother.  
[01:20:03] >> Oh, so let me repeat that definition. So we say violence against women because we categorize you as a VAW father. If at any point in the data in which you see you we see you you have been suspected of violence against a woman. This could have happened before the child is born  
[01:20:17] we we can show in robustness checks that if we restrict to years where the child was 0 to 18 the results all stay. You know what I'm trying to say is like but from a child's perspective being exposed to that guy I am exposed to specifically  
[01:20:30] IPV portion of it if it's a sorry in the work I'm not able to get exposed to it but I'm just exposed to he's beating up my uh like like isn't the treatment is more like I'm exposed  
[01:20:43] to a father yeah it's violent but like I am exposed to the home version of it.  
[01:20:48] >> Yeah. So what I will say is with the Swedish data we cannot see the victim ID. Now, most likely, this is why we do robustness checks from it happening 0 to 18. Most likely, if it's 0 to 18 and it's violence against a woman, there's a probably a high likelihood it is the mother in the household, but not for  
[01:21:02] sure, and we can't know for sure in our data. So, that's why we're we're very loose. Now, what I would say is if this ever suspected of VIW is not actually translating to modeling the home, then that's another reason why we would find no effect when we remove the father. And  
[01:21:16] what I would say is even if you are only ever suspected before your child is born, you could continue to be abusive and it just never be reported. And so, you know, we can't, you know, obviously we can't see things that aren't reported just like Sarah was saying, but we there is a kind of natural test of this, which  
[01:21:30] is if none of this behavior is being modeled in your house, misogyny, violence, etc., then removing the father shouldn't matter. And so the results are going to speak for themselves on that.  
[01:21:39] Yep. small psych thing but you know to test the innate trait theory do you see anything on woman you know daughters perpetrating violent crime >> so when we don't have I don't believe  
[01:21:51] the correlary of so it's hard to measure in our data women >> but we we do look at women violent crime perpetration in the appendix table that I'm I'm going to skip but I'll let I'll let you guys come see it after the talk um because we look at the women's side  
[01:22:05] as well there um but it's but women in general um just don't come in as violence. Sorry guys, as we >> violence against >> Yes. Yes. We could look at child abuse for the for the daughters. That's for sure. All right. So, we have an empirical challenge though, which is we  
[01:22:19] want to both the mechanisms are murky. It could go either way. Empirically, it's also hard to estimate because women who voluntarily lo leave a VAW man, potentially one who is abusive to her, are probably very different than women who don't. So, for example, they might  
[01:22:33] have much greater social support. They haven't been as isolated from their family who helps them leave. They might have more confidence. they might have better outside economic opportunities.  
[01:22:41] All of those things can separately help children have better outcomes in adulthood as well. And so what we're going to do as our main specification is try and find a way to quasi randomly remove fathers from the home. And then I'm going to show you two other ways to verify this main specification and  
[01:22:56] provide external validity because I'm always much more convinced when I can show you a result three different ways as opposed to just using one methodology. So let's start with the main uh most internally valid most causal I think approach we have which is we're going to take all of these  
[01:23:10] violence against women men we're going to restrict to those who show up in court for any crime. It doesn't necessarily have to be violence against women. I'll come back to those sub cases in a little bit but men who show up for any crime and then some of them show up to a harsher judge and are quasi randomly removed. Some of them have a  
[01:23:25] more lenient judge and are not. Now we've all seen lots of judge IV papers.  
[01:23:29] Um, I promised myself I would never write one again. Um, cuz I had to throw away two cuz they didn't survive enough robustness checks and I didn't believe them. Um, this one did. So, I'm here I am. So, the second stage is we're looking at whether the son perpetrates  
[01:23:44] violence against women as an adult or whether the daughter cohabits with a partner who commits violence against women as an, you know, as an later in life. And we're going to instrument that with whether the father is removed and goes to prison. And I'm going to talk about kind of a first stage about how  
[01:23:57] exposed they are in a second. But for now, right hand side is dad is quasi randomly sent to prison and what happens to son perpetration of VAW and daughter victimization risk. We're going to instrument whether the dad goes to prison with the the randomized judge  
[01:24:11] they're assigned to. We have a nice first stage. A 10 percentage point increase in judge trenyy translates to a 5 percentage point higher likelihood of the father going to prison. Yep. Jessica  
[01:24:22] >> in Sweden are people are judges um separate is there like a separate domestic violence court >> so yes and so we do it conditional on court by year by crime age strata because yeah to deal with that so  
[01:24:36] there's a whole long appendex looking at the judge IV it's not quite as long yet in the paper I posted because that's very much work in progress um but there will be a longer one coming too um but yes that's really important because there you know judges do specialize um so that's been true of lots of judge  
[01:24:50] papers if you look at them. Usually it's crime court by year uh crime strateg strata which is how the randomization happens. Um yep.  
[01:25:00] Just a quick question about um the the guys who are going to uh to prison because they perpetrated a crime uh against women uh versus any other crime.  
[01:25:09] Is there a difference between those two like groups?  
[01:25:11] >> I will get to that in one second. So for now we're looking at a guy who goes to prison for any crime, not necessarily violence against women. I'm going to talk about the subset in a second. All right, so we have balance. Um, I'm going to go quickly because I have two more identification strategies after this.  
[01:25:25] But the first question, and I give Janet Curry a lot of credit for this because she told us to look at this and it was brilliant because she's brilliant obviously. Um, so there's a common frame that nothing works when it comes to getting women to separate from abusive partners. And I think that's become less of a refrain recently, which is good,  
[01:25:40] but there used to be a much stronger drum beat of this. And so one question that's really important in the Swedish context is we're not in the US where you go for jail for years. We're in Sweden where the medium prison sentence is 3 months. That's not going to lead to  
[01:25:53] long-term less exposure for sons and daughters if there's not a durable separation. And I think this is actually a critically important policy question because is giving women just a little bit of breathing room enough for them to facilitate a longerterm separation from  
[01:26:07] a pretty bad partner. All right. And so what we find is that's true. So we find that the probability that the child lives with their bio violence against woman dad drops by 73 percentage points 5 years after trial. So these three  
[01:26:21] month prison sentences give you enough breathing room if you're a woman who experiences a relationship with a violence against woman man to affect a durable separation. Now you might worry, well is she just cohabiting with some new bad guy, some new violent guy who's  
[01:26:35] violence against women? And no, a little bit. So this goes down. But the probability that the mom cohabits with any violence against woman man drops by 47 percentage points compared to the women whose uh partners were not taken  
[01:26:48] away by the slightly harsher judge. All right, for 3 months. Now the last thing to directly speak to the woman's environment is does the woman is she less likely to show up for IPV care and the healthcare hospitalization data? Do we see actually that this reduction in  
[01:27:02] partnership with violent men leads to less abuse that she experiences? And again, this is a rare outcome. So we don't have a lot of precision but this is significant at the 12% level and we do see a 12% you know level confidence interval but we do see a reduction in  
[01:27:16] her IPV experience. Now this is our main outcome. Yep. Jessica >> does do anything except for the judge.  
[01:27:33] [snorts] >> Yeah. So this is the common exclusion restriction. So on the one hand I'm in some ways I'm less worried because that bundle treatment seems to be working but on the other hand to to take it to other countries we would like to know more about what's in that bundle treatment.  
[01:27:47] So we will definitely have on our to-do list is look at everything else we can see in the admin data to get at what else might be in there. Now, for the most part, we think the there's less bundling of punishments um than there is say in the Finnish context um from from  
[01:28:01] what I remember, but I we will we we will expand on that section um because it's really important if we want to think about with this work in the US, for example, is there something special they're doing in Sweden that we should be doing in the US when we when we intervene with women who are experiencing violence in the home, for example,  
[01:28:15] >> particularly bundling of punishment with resources.  
[01:28:18] >> Exactly. Yeah. Yeah. Yeah. And so you might imagine this is why I will say we cannot do the judge IV in the US. We are interested in doing a matched court version of this and I I'd love to talk after the talk about what how credible people think that is because the US is a very different context in terms of  
[01:28:32] support. on this question. Does it affect the custody arrangements afterwards? Like will the father the kid in any >> we have checked this and I forget the exact numbers but we can see where the child is residing and there are cases  
[01:28:46] where just the the child does get removed from the home entirely but and I can't remember the numbers off my top of my head but that will be in the paper like how often they're residing with the mother after you know so most moms are not partnering or so most children are not living with the VW dad. Most of  
[01:29:00] these cases they're living with the mom but I don't remember the exact number.  
[01:29:03] like moving is one. [snorts] >> Absolutely.  
[01:29:10] >> Yeah. So that's one thing we are interested in looking at if he moves municipalities and we just haven't had time to code the municipality also for my next identification strategy. It's less of a identification strategy than a descriptive exercise but we haven't we didn't have time to code it for the summer institute. We tried very hard but  
[01:29:24] we had other things. Let me pause and or I'll take one more.  
[01:29:27] >> Sorry last question. Um so two things. One is that you know u the father actually getting punished by a judge or being hand did a sentence. It may also show that you know these crimes yeah there's deterren >> I'm going to get to that in one slide.  
[01:29:42] So let me get there. Yeah. So let me look at my main result first. Um so what do we find for the kids outcomes? If your father is quasi randomly removed because of a prison sentence, you as a son are 533 percentage points less likely to commit VAW. Now that's a huge  
[01:29:55] number relative to the complier mean. It's not over the complimentary mean, but it basically almost eliminates the behavior. This was an astonishingly large result. Now, I would say the confidence intervals are not tiny. That being said, these are the extremely bad guys. These are guys who when their  
[01:30:10] children are getting older are still committing crimes that land them in court. And then there of the compliers who are in the margin of going to prison, which in Sweden puts you on like the very right tail of bad guys. So I think removing the worst guys seems to  
[01:30:23] almost eliminate this behavior amongst their sons which is an astonishing result to me. Um even if we take the bottom of this confidence interval this is still a large and good you know good effect in terms of reducing violence against women. We also see a reduction in victimization risk for daughters.  
[01:30:38] They are 65 percentage points less likely to cohabit with a man who is suspected of violence against women. And again that almost eliminates um 82% decline relative to the control complier me. These are such big results. I want to just strongly stress that these are the worst guys. I only have 11 minutes  
[01:30:52] so I'm going to push ahead a little bit because I want to quickly talk about mechanisms. We have this bundle treatment of you're getting less of this modeling or misogyny or violence exposure. You are also potentially having better maternal outcomes. I don't have time to show this. I will say we  
[01:31:06] find that women are mothers are less not only are they less likely to show up for IPV flags, they're also less likely to show up for alcohol addiction. And so it seems like the mothers have a recovery after their uh bad partners are removed.  
[01:31:17] For deterrence, I think there's something very interesting. We find no deterrent effect for boys. All right?  
[01:31:21] There's no difference in these estimates if we just exclude VAW. Now, prison sentences in general might be deterrent, but VAW prison sentence in particular might be deterrent. Someone asked about deterrence for daughters. On the other hand, there seems to be a strong deterrent effect. Seeing your father be  
[01:31:36] incarcerated for specifically violence against women, if we exclude those, that does away with the effect. And so there seems to be some sort of deterrence mechanism for the daughters at play. Now I want to go to two additional identification strategies that to us is our really strong causal internally  
[01:31:50] valid estimate. I like to see things multiple ways to be convinced of results. I would also say that there's a concern about external validity with that sample. That sample is extremely negative selected not only to other men in Finland but to other men who are  
[01:32:04] suspected of violence against women. They are so just look at the sun's uh interracial transmission at VW. it's much larger. These men who end up in the judge IV sample out of the men in the BAW sample, they earn less. They are in  
[01:32:17] more crimeridden communities, etc. So, you might worry to what extent did this estimate, this very strong causal estimate I've shown you, to what extent would this, you know, extrapolate to broader families who are less horrifically  
[01:32:31] I shouldn't make such value judgments, but less criminal fathers, let's say, um, so having committed a VAW, but aren't still committing crimes when their child is 8 n 10. So to look at this we're going to do two additional things. Number one is we're going to look at broadly the influence of male  
[01:32:45] role models descriptively and then we're going to turn to a family fixed effects model to look at it a little more causally. All right. So first and this relates to one of the questions that was already asked. Let's look at our whole sample. So this is just that LPM  
[01:32:59] estimate we started with. 11 percentage points more likely to be suspected of VAW if you grew up with a VAW dad. 8.5% points more likely to go habit with a VAW partner. If we look at those who never actually go habit with their VAW  
[01:33:13] dad, that correlation almost halves. If we further say okay well let's take the sons who never cohabit with the biod or sorry who in column three who have a stepfather so some other male role model  
[01:33:26] most of whom are not v dads if it drops even further. And last, if we look at cases where you were adopted by your stepfather, which is usually a combination of you were, you know, never ex living with your biod and you were adopted by some other male role model,  
[01:33:40] that interracial transmission goes away entirely. And so what this says is it's suggestive of maybe two things. Number one, if you reduce the exposure of VW dad, even in these less severe cases where they're not going to prison, we still see a reduction in that  
[01:33:54] transmission. And we see the same reduction, although it doesn't entirely go away for daughters. Number two, and this is someone asked about other male role models, we can do something even more interesting and say, okay, because I often present this to a room of mostly men, and then they get very worried  
[01:34:08] about, are you telling us we should remove men from the family? And what I would say in this regression is we can say okay well for sons what happens if you replace so if you look at say sons who grow up with a a biod who was violent against women what if we bring  
[01:34:22] in a dad who has never been suspected of v and we see with this negative coefficient that it has a pretty strong mediating effect. It actually in this potentially good male role model decreases the transmission of violence against women. And so I think this is  
[01:34:37] again this is suggestive evidence that good male role models can actually mediate this. They can actually provide undo some of this transmission about how you treat women as adults. Bad uh male role models on the other end here if  
[01:34:49] even if you're your biological father never committed BAW, if you bring in a stepfather or another dad figure in the household who cohabits with you who does commit BAW, then you become more likely to do it yourself. And so this table and  
[01:35:02] this is just raw data. And I I sometimes I really like showing just raw data because it seems to really speak to this male role model effect. Bringing in a violent dad um both if you've never been exposed before or if you've exposed before increases the transmission.  
[01:35:15] Bringing in a non-violent dad decreases transmission.  
[01:35:19] Now for daughters we don't find quite as stark of a divide. We find that having more so we find a really strong correlation between bringing in any VAW male model male role model and partnership risk. We do find not  
[01:35:32] significant here for the nonVW dad, but positive and positive and very small and significant. But what I will say is daughters who are exposed to more male role models are more likely to have more partners themselves. And so we think part of this is mechanical. You just end  
[01:35:45] up having more partners as well. So that seems to be a norm that transmits.  
[01:35:50] And what I'd like to close with in my last few minutes before I conclude is our final piece of evidence. So of these three pieces of evidence to show that, you know, like Karen was saying, we have these really strong descriptives of transmission. We've shown you if you send you to prison, it reduces the  
[01:36:04] transmission by a lot for these really bad guys. So if you take the worst guys out, reduces a lot. With this descriptives, we find it also we find consistent evidence. Now, we're going to try and take that and go a step further.  
[01:36:14] We're going to do a family fixed effects design. And so if you like Rajett's move to opportunity work, if you liked that great divorce paper, we're going to take the exact same identification strategy and take two siblings. one of and then this is voluntary separation so there  
[01:36:28] there are some indogenity concerns there but we're going to look within sibling which hopefully gets rid of some of those concerns and we're gonna say do you have more or less exposure to this VAW dad and does that change transmission all right and so in our main specification we're not going to have  
[01:36:43] this second line I'm going to talk about this in a second and the first set of estimates I'll show you ignore this second line we're going to look at the years of cohabitation and we're going to have these family fixed effects cohort fixed effects and some controls for father birth year kind Going to your point, Claudia, there's different  
[01:36:57] violence across different cohorts um and and other other controls. And I'll talk about this other coefficient in a second. And what do we find once again?  
[01:37:07] We find that the longer you are exposed to your violence against women father, the more likely you are to perpetrate violent against women yourself. And if you aggregate this up, what's really interesting is it is a significant effect. It is much smaller than our  
[01:37:22] judge IB effect. And to me, this makes a lot of sense. We're saying that like 10 years of more exposure gives you a 2 percentage point decline roughly in violence against women. All right. Now, that relative to our baseline is, you know, a 20% about decline in the  
[01:37:35] coefficient, but it's not the entire thing going away like with the prison dads. I think the difference here is we're taking the worst worst guys out.  
[01:37:42] Usually pretty early. And I didn't have time to show age he hogenity, but for especially for boys, it matters that it's super early. So, if you take these super bad guys out, then you can almost undo it. For most dads, I do think you have this combination of effects where yes, these are guys who have been suspected of violence against women, but  
[01:37:56] they're also probably doing some other things that are good for their kids. And so you get this smaller coefficient size for these guys who aren't showing up in court over and over again. Now, on column two, again, for the men in the in the room, if you are not a VAW dad, all  
[01:38:10] right, so if we look at families where there the dad is never suspected of violence against women, we don't see that transmission. And if anything for daughters, having the dad removed if the dad never committed violence against women actually increases partnership  
[01:38:24] risk in the future. So that's kind of consistent with that divorce paper, consistent with other papers that have shown that usually removing dads results in worse outcome for kids. But when you remove dads who are providing a role model of violence against women, it  
[01:38:37] actually yields better outcomes in terms of partnerships for women and perpetration for men. Well, one thing is is the other side of this which is the push back from women and outside options which you can add in.  
[01:38:51] >> Oh, yeah. We could do like a bar instrument or something.  
[01:38:54] >> Yeah. Something where this this is only sort of one side would just plop this bad guy on this woman, but [snorts] she could h, you know, have some push back and push back that might grow over time, which is in some sense what you're  
[01:39:08] showing with the three months without him. Perhaps she gained skills or gained confidence.  
[01:39:15] >> I think a great thing we could do is replicate Anna Eiser's outside opportunity results. See if to the degree to that which that leads to separation. Do we then also see the perpetration decrease? I love that we were able to replicate that at least using an employment instrument in Finland. Earnings there's not enough variability with an occupation to do the  
[01:39:30] Bartick, but we could probably do something similar in this week. I love that. So kind of other other ways that help the woman indulgely choose to leave a partner who is not helping her succeed in life and and and potentially being very abusive and harmful to her. All  
[01:39:44] right. So in my last uh two minutes to leave a little room for questions I want to conclude. Um so what we've done in this paper is in the first part of the paper we established the first time using large scale administrative data and we will be adding US correlations as  
[01:39:57] well. We establish that there is a very large transmission of violence against women from father to son or at least there's a correlation. Uh let's be more muted here. But hopefully now Karina I convinced you that in the next part of the paper we showed you using three  
[01:40:12] different strategies that actually if you remove these fathers who are violent from the home who who ex who have ever committed violence against women and are likely exhibiting those behaviors or those norms to their child we see a very strong reduction and violence against  
[01:40:26] women perpetrated by sons and experienced by daughters. To me despite this being a very depressing paper this is where I hope to go. This is where I think we can go as a society is figure out what causes this behavior and then figure out ways to reduce it. And and what we're showing here is not only do  
[01:40:40] male role models who exhibit violence against women behaviors increase transmission, but good male role models can reduce it. And so the more we can get good more role models into young men's lives, I think the better we outcomes we will have in terms of women's outcomes choosing partners who  
[01:40:55] are less abusive towards them and also men choosing to treat women in a way that is not uh rife with violence. And so I'd like to close there and thank you so much for the excellent comments.  
[01:41:08] [applause] Yeah, I just wanted to say so you're discussing this in terms of learning, right? You're learning role model and so  
[01:41:22] forth, right? But this is could be there is a process of scaring of PTSD. So I don't know, you know, this could be a learning mechanism in the sense you reproduce behavior >> or it could be that you are damaged  
[01:41:37] somewhat seriously for the kids. You see what I'm saying? And that's the it's slightly different to me.  
[01:41:43] >> Yeah. where are the two concepts right you know these kids are traumatized >> yes absolutely >> boys or girls so I don't know you know how do we is it just learning is it preferences >> one thing I will say that's reassuring is I think this speaks to our family  
[01:41:58] fixed effects design which is okay the longer exposure probably the more traumatized you are but I guess a really sad note would be if a child is only exposed say four years are they indelibly traumatized and they're never going to be able to recover and the  
[01:42:12] family fixed effects model says No, it says that actually we can undo some of these harms. Now, those coefficients are smaller probably for lots of reasons.  
[01:42:18] One of which may be like, you know, we're not getting like cuz so I would say with the for the judge IV design for daughters, it's kind of big across the board across ages, although we're not powered to really do heterogeneity by age, but for sons, it's concentrated in young ages. Um, but I think the family  
[01:42:33] fixed says it's not irrevocable the the trauma, which is good because if you, if you've done any reading, and I don't work in this area, but I've done a lot of reading of looking at trauma of kids and foster care and so on. Um, it, you know, the one of the terms of phrase for foster care is like it takes, you know,  
[01:42:47] triple the I forget what it is triple the time out versus the time in to undo the trauma. And with the family fixed effects, it seems like, well, we can undo it. There's going to be that trauma there. And the other thing I would say just broader from a mechanism's point of view. The goal of our paper was to say does the correlation exist. Can it be  
[01:43:02] cut? And then the extent to which we can understand more why is really helpful.  
[01:43:05] But to some extent this is always going to be a little bit of a bundle treatment because we do show moms do better um which also helps with parenting inputs.  
[01:43:12] Trauma is potentially undone by having better better caring environments. And then on top of that there's a learning that's not transmitted. It's I'd love any ideas you have to even further disentangle those three. Those are very hard to disentangle. Uh but I I think  
[01:43:26] this is a really good starting point in terms of what we can do.  
[01:43:30] >> One thing we look at teenage years if anything there's improvement actually.  
[01:43:35] >> Yeah. >> Yeah. Yeah. So mental health improves education does not. So we don't see like big education outcome.  
[01:43:40] >> So we look at for example from removing the >> Yeah. Yeah. So the trauma seems to go away.  
[01:43:49] >> Question. >> Yeah. Yeah. Yeah. We should be careful on that.  
[01:43:52] >> Okay. So, we're going to take a break for uh You should