Episode 85: Sofia Amaral

 

Sofia Amaral

Sofia Amaral is an Economist at the Center for Labor Economics of the ifo Institute at the Ludwig Maximilian University of Munich.

Date: December 20, 2022

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Dr. Amaral's work on sexual harassment in public places and what to do about it:

“Sexual Harassment in Public Spheres and Police Patrolling: Experimental Evidence from Urban India” by Sofia Amaral, Girija Borker, Nathan Fiala, Anjani Kumar, Nishith Prakash, and Maria Micaela Sviatschi.

[Working paper available from the authors.]

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 Doley at Texas A&M University, where I'm an economics professor and the director of the Justice Tech Lab. My guest this week is Sofia Amaral. Sofia is an economist at the Center for Labor Economics at the EPO Institute at the Ludwig Maximilian University of Munich. Sofia, welcome to the show.

Sofia [00:00:31] Hi, everyone. Hi. Thank you, Jen, for inviting me.

Jennifer [00:00:34] Today, we're going to talk about your research on sexual harassment in public spaces and what to do about it, but before we get into that, could you tell us about your research expertize and how you became interested in this topic?

Sofia [00:00:47] Sure. So I've been working almost exclusively on the topic of gender based violence and violence against children in a variety of contexts, such as in India, which we're going to talk about today, but also the UK, Mozambique or El Salvador and the main reason why I got into this is because first, it is a very huge quality problem. The rate of victimization of the different types of offenses within gender based violence are very high. Of course, we know that one in three women are victims of domestic violence, but we also have, for example, when it comes to sexual harassment in public spaces the topic of today, about 50% of women worldwide have experienced it and the truth is that from a research perspective, we know very little about it for a variety of reasons and this is what got me really interested in addressing this problem and trying to think about ways to contribute to the discussion.

Jennifer [00:01:50] So your paper is titled "Sexual Harassment in Public Spaces and Police Patrolling: experimental Evidence from Urban India." It's coauthored with Girija Borker, Nathan Fiala, Anjani Kumar, Nisith Prakash and Mica Sviatschi. Quite a dream team there. So set the stage for us a bit in terms of the context in India specifically, what types of harassment are you interested in and how big a problem is this in that country?

Sofia [00:02:16] Yes, a great team. I'm very happy to meet such knowledgeable coauthors. So street sexual harassment in public spaces, in urban settings in India is a very big problem. So we know from recent surveys that about 79% of women are subject to it on a yearly basis. We also know that in our city, in Hyderabad, the city where we work in, 75% of women feel unsafe after 4 p.m, so 75%, this is a huge rate of feeling unsafe. We also know that 87% of these women take some measures to protect themselves from sexual harassment.

Sofia [00:03:10] So here we're seeing two or three things. One, that harassment, victimization of harassment is very high, that women form perceptions about it throughout their day, and that they change their behaviors creating coping mechanisms to protect themselves from sexual harassment. And the question is, then, how does this impact women's entry into the labor market, women's physical mobility and well-being while they're out and about in participating in the economy? So this is what got us really interested in why this is such a big problem in India, but also in either urban settings in developed or high income countries.

Jennifer [00:03:56] Can you give us some examples of what kind of harassment we're talking about here? Is this like people staring or calling out at you or what should we have in mind?

Sofia [00:04:05] So sexual harassment, so in the literature is defined as being forms of unwelcome sexual advances that involve verbal, nonverbal or physical. So it's a very wide spectrum of severity that can go from staring to stalking to intimidating to touching, groping or taking pictures without consent.

Sofia [00:04:31] So to the more severe but also more rare events like rape or physical abuse. So there are these very different types of harassment and this is actually going to be important in our paper, too, distinguishing very well between mild forms of harassment for which the rates are higher, but are also less consequential to more severe forms of harassment that are more consequential but also have lower rates of victimization.

Jennifer [00:05:03] And so you mentioned that we don't know much about this from a research perspective, so when you guys were first starting this study, what had we known? What evidence was already out there about how to address this problem?

Jennifer [00:05:16] Yes. So there's two sets of literature here. On the one hand, we have some work in sociology describing the problem by its incidence, describing this aspect, trying to really understand how to define and how to think about harassment. So they did very interesting work in establishing the pattern is that harassment is more likely towards women, it's more likely committed by strangers, there's this very widespread jump of severity of offenses. So this is the first piece of research that has been out there for some time now. And then when it comes to research in economics, our discipline, we have work from my coauthor, Girija Borker showing that indeed women make very rational decisions, for example, when it comes to choosing university college, depending on how safe their commute is and they're indeed showing that they choose colleges of lower quality that they could have gotten into in order to have a safer commute. So here we have a really important piece of evidence showing that there is this relationship between safety and a very tangible economic outcome. We have the work from Zahra Siddique looking at the correlation between perceptions about harassment, where you reside, harassment towards women, and how it lowers female labor participation.

Sofia [00:06:51] And this again, goes back to this point that sexual harassment in perception is about to have a negative correlation in women's economic outcomes. And we have also evidence from Brazil, from funding leads showing that women are actually willing to pay for a safer commute and this seems like a very trivial result, but it's not so trivial because it is putting a number, a dollar amount to how much women really care about their safety and how much they really integrate safety into their economic decisions on a daily basis. So these are three pieces of evidence, really highlights how important this is for women's economic empowerment.

Sofia [00:07:34] However, then the issue is that there is a missing link establishing a direct causal relationship between safety and women's economic outcomes and the reason why this has been a challenge in the literature, because it's really difficult to change safety. How do we vary safety of sexual harassment in this context is of large, complex urban spaces, and this is where I think our work really makes a contribution.

Jennifer [00:08:07] And so what makes this so difficult to study? What do you see as the main hurdles to figuring out what works to reduce this type of sexual harassment is mostly a data challenge, mostly identification challenge or both of those things, right?

Sofia [00:08:22] So I think data is the first big challenge. One, of course, we cannot use the reporting data and here it's really an absolute no no because this is hardly ever reported. I think in our surveys we see that only 60% of women have ever reached out to the police to discuss sexual harassment instants. So this is really not an option here. Two there is and much like any other form of gender based violence, a stigma to revealing to enumerators or trying to disclose this information when women are interviewed.

Sofia [00:09:04] And in the third point, which then makes it more difficult from a research point of view, is that it is a very frequent offense. So even if we were to survey women, say after a year in between that here, this happens so often that women have a difficulty recalling exactly how harassment took place in when and what time and what type, for example. So this makes it, of course, super challenging to measure in the first place, which is a necessary condition to study. And then from a research point of view and identification challenge, harassment is a bit of a byproduct of norms and probability of punishment in the sense that it is very tolerated by even victims, perpetrators, bystanders and state actors. So it really becomes difficult to understand, okay, how are we going to change the level of violence in this context and when we have this combination of high, tolerable, offense with very low probability of punishment.

Sofia [00:10:15] And there's also another issue which is a little bit difficult to visualize is the fact that harassment in our context in Hyderabad, but also in other large cities, so Rio de Janeiro or Mexico City, they're very complex urban spaces and where we don't really understand or we it's difficult to measure the jurisdiction, for example, around a bus stop or in a large square where people are commuting, so this makes it really difficult to pinpoint the geographical space where these things are happening. So all these data issues, but also in terms of identification make it really challenging. And it's also a little bit different from when we think about studying domestic violence or studying sexual harassment in the workplace, because there we have a physical set space where things happen in a tangible relationship where this type of problem exists, but here it's not so much like this. So this is the challenge.

Jennifer [00:11:24] Okay. So you worked with the Hyderabad City Police to study the effects of a novel policing program called "SHE Teams." So tell us about this program. When was it created and what does it involve?

Sofia [00:11:38] Yeah. So "SHE Teams" is a really fun and really well chosen name for this task force, actually. It stands for Safety, Health and Environment Teams. It is a specialized task force that only handles sexual harassment in public space. This is their task. They are not dealing with traffic, they are not dealing with robberies, they're not specializing in any of these. This force is a it is very novel, at least we haven't heard in our group any type of policy of this kind. They started operating in 2014 and remember that they India, of course, cities are often very highly populated. And in this context, this is not no joke because they serve 9.7 million people. So it becomes a very interesting context to study in the first place.

Sofia [00:12:34] And what they do exactly is so they initially they patrolled undercover and the idea is to really focus on arrests, detecting perpetrators, identifying them. They created a criminal history for these perpetrators and monitor them. They talk to their families and they receive fines or are arrested and taken to court, depending on the severity of the offense. They also have very high social media presence and the idea behind this undercover policing was really to play with two aspects. One, that there is is a relatively small task force for the scale of the problem and two they want to create this

omnipresence of the police around the city, so have a sense create a sense, too, on victim, on women and on potential perpetrators that the police is there. You don't really know where they are, but they are around. So be aware that you're being monitored for this purpose of sexual harassment. So this was the idea.

Jennifer [00:13:43] And so they implemented that before you all got involved, or was that were you involved in the start?

Sofia [00:13:49] Yes. So they were operating at a very small scale, but already there was quite a lot of interest from the media, from researchers and policymakers, because there's really not much being done anytime in India on this topic.

Jennifer [00:14:07] Okay. So once you all got involved, this program expanded a bit, I gather. And so then it moved on from just being this undercover aspect to having uniformed police as well. Right. So what were the teams that that you all studied?

Sofia [00:14:24] Right. So when we first started engaging with the police and starting understanding the degree of change behind the type of police patrols, we got into a very interesting discussion and in the end, we agreed that it would be really interesting to vary two things in terms of the police patrolling. One, let's keep the status quo of undercover policing and in any case, we're improving the presence or increasing the presence of police in the streets of Hyderabad, but we also wanted to test this type of police patrol on a belief that is visible in which the officers are in uniform. And so we're varying this both the presence, but also the visibility and this is really important in a paper because of the following. So if we think about undercover policing, they are right if a police force is maximizing arrests, then undercover is probably really efficient because you get the element of surprise when you're trying to identify perpetrators committing an offense in the act.

Sofia [00:15:30] This all seems really reasonable, but harassment happens at a very large scale in the city. So we thought, okay, but maybe we know from the literature that police patrols work and they work when they are visible because we think they have a much more higher effect in terms of the daring as opposed to incapacitating individuals. So let's incorporate this and for harassment is becoming a really important part of the story because it's really signaling to victims and perpetrators as well that the police is there and cares about this program that affects women. And the way the way we do this is to really reform teams, each team has three officers, they have to have at least one female and they follow a typical police patrol program of visiting different hotspots throughout the week.

Jennifer [00:16:25] Okay, great. And so you've touched on this already a little bit, but just to walk through it kind of in a bit more detail. So what are the various ways that these "SHE Teams" both the undercover and the uniformed teams, might affect sexual harassment in public spaces? What are the mechanisms that we're most interested in here.

Sofia [00:16:44] Right. Yes, exactly. So we think or initially when we before running the study, we thought that their their presence could change victims behaviors. They feel women would feel more empowered to be more out in the street, increased presence of women in the streets, but also at the same time could deter perpetrators by signaling that now we're really that you are being identified if you commit such a crime is higher and therefore you're going to update your beliefs about punishment and therefore you won't

commit such offenses. So we would have these two sort of mechanisms here, at least on the women's side, so for the victims and on potential perpetrators. So this was a little bit the idea.

Jennifer [00:17:38] Great. And then you mentioned that you're going to randomly assign these patrols to different hotspots. So you're running a randomized controlled trial here, a nice field experiment. So how did you decide where the hotspots were or where to send these different teams?

Sofia [00:17:54] Yes. So what we did before starting is two things. So first, we collect a survey of women who are commuting around areas that the police had identified to be areas of high sexual harassment. We interviewed about 230 women per area before starting everything. And in there we asked women about their experiences with harassment in the previous months. So this was the first exercise. The second exercise that we did was to recruit enumerators and train them to be as like the police and almost patrol these areas and code which forms of harassment they were observing while they were at the hot spot.

Sofia [00:18:43] So suppose I am one of these enumerators. I would go to Times Square, for example. Times Square is very large, but I would go to Times Square. I would be there for 20 minutes observing at about five women, and I would code what is there in the victimization of harassment that they are facing, how they look like, what type of harassment are they facing? So then what we did is to combine these two sources of data. One is observed so it doesn't have this aspect of reporting affecting the the measure and two is asking women directly where we think, well, maybe they want to tell us everything, but they are still going to tell us some very high valuable pieces of information that we can use. We combine this data in order to see when we might.

Sofia [00:19:37] This is that we have indeed areas, very clear areas across Hyderabad where these rates are indeed much higher than others we call these hotspots. So based on this, we've identified a variety of these hotspots. And then we worked with the hotspots, identified for the study to be areas where harassment is very high. And then, of course, as you've mentioned, we can't just allocate police to these areas, we'd add any source of variation because of course police placement could be correlated with the characteristics of the area. So indeed what we ended up using as a tool for identification is to vary randomly where officers would patrol and how they would patrol either in uniform or undercover.

Jennifer [00:20:33] And then how do you use that, that randomized experiment to measure the causal effects of the "SHE Teams"?

Sofia [00:20:41] Yes. So these officers, they were patrolling for about 24 weeks. So every week they would go randomly for short periods of time to an area. They would conduct their activities of patrolling and we collect this information of where teams of patrols are at each point in time. So we actually have very good data collected through GPS tracking devices that were in each police vehicle, so we track where officers are. At the same time through during the 24 weeks we have this exercise of women observing what is happening in terms of harassment at each of these points in time and remember, this is really crucial to hear because as I mentioned before, this way we have a measure that is of high frequency that is not related to the intervention, because the enumerators weren't aware that there was this underlying RCT going around. So they are blinded to the actual

experiment and they are not the victims themselves, so they are just telling us what they are observing.

Sofia [00:21:54] So we do our hitting right on the difficult challenges of harassment when we do this. And then, of course, because we have this very nice randomized experiment, we can really be able to identify the effect of increased presence of police in uniform in comparison to a control area where we have hotspot that are areas of also very high rates of harassment, but for which they did not receive any form of patrolling, whether in uniform or undercover. It's just business as usual and if we compare how this harassment in this control area relates to the two treatment arms that we have.

Jennifer [00:22:36] Great to see what the randomized controlled trials are very difficult to set up, but then once you've got them running, you just have to compare the outcomes across places and it becomes very straightforward to measure the causal effects, so wonderful. I'm curious, I would love to hear a little bit more background about how this partnership came about. How did you all start working with this police department? Did they come to you? Did you go to them? How excited were they about this? Did they take some convincing? Give us a little bit of the backstory here.

Sofia [00:23:07] Sure. So as I you mentioned at the start, yes, it is a very good team of coauthors, in particular, Nishith Prakash. He was brought up with a father that is a very senior police officer, so he is familiar with many senior police officers in India and he knew Anjani Kumar, which is great Anjani Kumar was the commissioner of Hyderabad at the time. So this became a very good friendship and helped started this conversation about how a researcher and policy maker meet up, what can we do? What are the problems they are facing? How can I help? So this is where things started.

Sofia [00:23:50] And then from then onwards, both Mika, Girija and myself, Nathan, we all went various times to Hyderabad to one build a relationship and to establish confidence in us about what we were going to do, how we were going to use the data from the police, how we were going to sit down with them and understand how the organization works, how the program works. This is very important. This was very important to them, but it actually was also extremely important to us because without it, we wouldn't be able to build an instrument of how to manage sexual harassment. What are the things we need to think about as victims, as perpetrators, as police officers when we're studying this problem, so this initial phase that we had of about 6 to 8 months, so it is a long time of building this relationship and establishing a common ground to work was really important. And from then onwards, once we had an agreement of what we wanted, where we wanted to go and what were the things we're willing to compromise and not really to compromise, and we just moved forward.

Jennifer [00:25:06] Amazing. Very cool. You have mentioned this data a few times, and yet the measuring harassment is is certainly not easy, especially in this setting. You can't simply rely on official reports because it's rarely reported to police. So you all developed this, this new observation based measure of harassment where you had these enumerators go around and and actually watch these areas. So tell us more about how that worked. How did they actually do this in practice?

Sofia [00:25:35] So we recruited about 170 women. They were working in batches and they were trained much like the police. So we were always trying to mimic what the police does, but with our enumerators, we trained them in understanding what is an instance of harassment of, for example, catcalling. If you see someone being catcalled, yes, that is an

instance of harassment. If you see someone being stolen their wallet, well, that is not harassment. It is an offense, but it is not harassment. If you see someone being stalked or being taken pictures and they are not really aware, they are not feeling comfortable about someone having their picture taken without their consent, that is the harassment. So all these different items, I think 14 items that we trained women in, they actually used theaters illustrating different types of harassment. They practice in a real setting. These are very complex spaces, how to identify these instances.

Sofia [00:26:48] And then what happened was very simple they were just regular women, just like you and me walking about in the streets citizens wouldn't be aware that this was happening at all because it's just like a regular person that is there, they would go to an area, they would be there between 15 to 20 minutes with their mobile phones. They would be observing what was happening. They would observe up to five women and they code if there had been a victim of harassment, yes or no? And if yes, what types of harassment were the victim of? Once this is finished, set and collect the data on their mobile phones and move on to the next location. And this happened for a period of 24 weeks. And of course, this is a really difficult job remember that you're thinking about harassment and victims of harassment your whole day and while walking from place to place, this is really difficult. So because of this, they were working in batches. So at a time we would only have about 20 women doing this, this task, and then a new batch of women would come to initiate the same task, this would avoid fatigue and any concerns relating to any potential drama of having to constantly be reviving what harassment is.

Jennifer [00:28:22] Yeah. And I imagine there you had you all had to think about ethical concerns here, too.

Sofia [00:28:27] Right.
Jennifer [00:28:27] I mean, if these enumerators are watching and observing harassment

and observing potentially women who are in trouble, what are they supposed to do?

Sofia [00:28:35] At the end of the day, they are women just like you and me and this would be the experience that they would have if they were in any other job. Two they would always had an additional person with them to help in case something happened and we also had a very good referral and immediate contact that they could use in case something happened. We also made sure that all our enumerators had a support system after every day in a field and after every week where they could talk about anything pertaining to the study. So here we we really tried as much as possible to be extremely careful and start thinking about the different ethical issues that would concern this task. Yes.

Jennifer [00:29:29] Yeah. Okay. So you've got this amazing data that these anywhere writers are collecting for you painstakingly. And then what other data did you have for this project?

Sofia [00:29:40] Yes. So another really important concern we had at that time was, okay, so we're going to ask the police to patrol in this area, but not this area. Why would they follow our instructions? How can we make sure that they respect the rules of the randomized control trial? How can we do this? So here each team of police patrols, they would travel on their own vehicle and they had this data was being tracked, so we know where they were at each point in time.

Sofia [00:30:13] So they would be complying or not complying with the shift that they were assigned, the route and the shift that they were assigned for that day. So this is really useful because then we can see exactly how many visits they did, how long did they stay there and exactly what they did when they stayed there. If they've seen a perpetrator in arrested this person, they granted warnings and this allows us to test one compliance with the rules of the intervention, two if there is a potential incapacitation effect from their presence in the hotspots. We alsocollected from the women's survey as I mentioned at the start, which is really good because then we can have measures of footfall and women's protective measures and baseline information about social norms around harassment.

Jennifer [00:31:09] Okay, great. And then, then so the main outcome measures you're interested in are what?

Sofia [00:31:14] Observed harassment by the enumerators and what type of harassment. So here, as I mentioned at the start in this topic, it's really important to distinguish between severe and mild forces of harassment because they are so distinct our main outcome is really the number of victims observed of harassment per exercise so this is our main outcome.

Jennifer [00:31:43] Okay. And then just briefly, what what is in each of those buckets, the severe forms of sexual harassment versus mild forms?

Sofia [00:31:51] Yes. So for severe forms of harassment, we have the physical abuse, intimidation, touching, groping, and for mild forms of harassment, we have catcalling, taking pictures without consent, staring.

Jennifer [00:32:09] Great. Okay. So let's talk about the results. What was the effect of "SHE Teams" on severe forms of sexual harassment?

Sofia [00:32:18] Okay. So first, so the belief was right. So indeed, undercover policing is more effective at sanctioning perpetrators.

Sofia [00:32:30] And we do find that in areas in hotspots where the police literally in uniform, they have a higher incapacitation effect when compared to uniformed police. However, both uniform or undercover policing, the rate of incapacitation is very small when compared to the rate of harassment that the enumerators were identifying in the in the streets. So, for example, we find that they are able to sanction about 7% of observed instances of harassment. So this is the first result.

Sofia [00:33:11] The second main result that we have is that undercover policing does not change any form of harassment. So there is no effect on patrolling undercover, on incidents of sexual harassment. However, uniformed policing reduces severe forms of street sexual harassment by 27%. So this is the first real piece of evidence linking changes in safety on street sexual harassment in public spaces.

Jennifer [00:33:47] So, so just to make sure I'm I'm following this so what you're finding here is that even though the undercover police were more able to make arrests and and get some of these these guys off the street, that might have had a small effect on sexual harassment, but overall, what you're finding is that the uniformed police had a much bigger effect on sexual harassment just through pure deterrence. Everyone saw the police were there and so they didn't do anything and they didn't need to be arrested in order to stop.

Sofia [00:34:17] Exactly. Exactly. So we do find evidence consistent with the presence of an incapacitation effect. However, this effect is too small to be able to explain the reduction in observed sexual harassment and simply because the rate of harassment is much higher than what the police is able to apprehend.

Jennifer [00:34:40] Mm hmm. So just having a cop who's standing there in uniform is much better than actually trying to get him to go around and and arrest people.

Sofia [00:34:50] Correct. Exactly. And then interestingly, linked to this, we find something really interesting, which is actually in areas that were being patrolled by uniformed police officers and in areas where harassment was reduced we also see a change in women's coping mechanisms associated with harassment. Particularly, women are less likely to be moving to other places, and they are also less likely to having to use bystanders in cases of victimization and this became really the first piece of evidence looking at how changes in safety by improving police presence in uniform, how this empowers women as measured by these changes in coping mechanisms that women don't need to have anymore because of this improved deterrence capacity of the police.

Jennifer [00:35:51] And so the uniform putting uniform police in particular. In these hot spots, reduced severe sexual harassment and allowed women to kind of change their behavior and stop having to protect themselves as much. What happened to milder forms of sexual harassment?

Sofia [00:36:08] So here it's really interesting. We see no changes at all, regardless of the type of policing for mild forms of harassment. So really, we this intervention of police patrolling, we're really just moving the more consequential types of harassment, the very severe ones. So the question then becomes then why is it that even with this form of patrolling that seems so effective for severe offenders, why are we not changing milder things? And this brings us to the next phase of this study.

Jennifer [00:36:46] Yeah. And then you you in addition to running this amazing field experiment, which is already a tremendous amount of work, you conducted a lab experiment with police officers. So, yeah. Why did you do this? And then what did the experiment entail?

Sofia [00:37:01] Yes. So we were really puzzled by this result that why is it that we're able to have a deterrent effect on severe form of offenses, but we're not seeing anything for mild offenses, what could be driving this? So this really took us into a very long discussion, including with the partners, to try to understand this. So then what we did was, okay, so maybe this is because police officers have different abilities, skills when it comes to detect sexual harassment. So I've told you, these are urban spaces, they are very crowded, harassment happens in very different forms. So could it be the case that their detection capacity is different by the type of offense within sexual harassment? This is our question number one.

Sofia [00:37:51] The second question was, okay, so maybe could it be the case that what if they think that doing something when it comes to severe forms of harassment is their duty, but when it comes to mild form of harassment, they actually are much more tolerable already and therefore they don't see a reduction in mild forms of harassment. So we had these two competing hypotheses. One, detection skills. It has a very clear policy action. Okay, let's train officers in being really good at detecting crime or okay, what if it's not

detection skills, but what if it is norms and officers tolerate different forms of harassment differently?

Sofia [00:38:39] So this is when we sat down, we built our lab in the headquarters of the police. We invited police officers that were in our intervention, as well as others, over 300 officers and we conducted a lab experiment where we show them different videos of different situations that could have happened in the streets of catcalling, of taking pictures without consent, physical abuse, groping, neutral offenses, property offenses and then we we showed these videos to officers in different ways and for each video, we asked them, what do you see in this video? Can you identify what is the main event in this video? Do you think there is a need to do something as a police officer with respect to this situation that is being depicted in movie in the video? And if so, what action would you take?

Sofia [00:39:43] So here we're really looking at measuring three things. One, our officers being able to detect what is happening in a video. So can they detect sexual harassment or not? Two, are they willing to do something about it. Are they willing to exert effort to apprehend a potential perpetrator? And if so, what punishment would they give? So these three essential components that would be embedded in a decision making of officers when they're out and about patrolling. So this was the idea.

Jennifer [00:40:21] And so what did you find?

Sofia [00:40:23] So we found that, wow, they are actually really good at detecting it. We vary the speed in which they they see the videos. We vary the types of offenses that they are seeing, different types of sexual harassment by severity and we do see that they have very high rates of detecting. So it's not the case that officers don't know what harassment is, they know it, so detection capacity is not the issue. However, when we compare their willingness to sanction and the actual punishment that they would be willing to exert in this experimental setting, of course, in sexual harassment cases, in comparison to property offenses. We do see that indeed they are less likely to think that there is a need for a police action and to give the correct punishment for the crime that is being depicted in each video. And particularly for mild offenses they are less likely to be willing to punish and they are punishment is less in comparison to property offenses.

Sofia [00:41:35] And when we compare mild versus severe forms of sexual harassment, we do find the same thing, that they are less likely to think that it needs to be punished and they are less likely to grant the correct level of punishment for that type of offense. So all in all, what this tells us and to put the puzzle together so we don't find evidence that they don't deter mild offenses because they don't know what it is. This is not the case and this was a really our first hint was this, but there is something on around norms associated with mild offenses that it's not there for other types of offenses. It is affecting job performance.

Jennifer [00:42:23] But you also found variation in that across officers. And so you went back and you compared how officers who behave differently in the lab, how they had behaved out in the real world when you ran your field experiment, were the real world effects of the "SHE Teams" correlated with what those officers in the lab.

Sofia [00:42:43] Right. So then we thought, okay, so we have this result in the lab, but we also have a really nice data where we know where each team, where each officer either did find in that case, let's combine this data and really try to understand even when teams have better norms, if they are actually able to address mild harassment. And we do see this, we do see that when patrol teams have better norms in relation to harassment, this is

a very specific to sexual harassment. This is really it's not general norms. It's norms associated with sexual harassment when they are more gender equal, more progressive in this dimension, they are able to deter mild forms of harassment when they're out patrolling. Otherwise, they are not and again consistent with the rest of the paper we do see that everything is again coming from uniform officers, not undercover. So we do have these component that one officers are able to incapacitate, but this is not enough for the scale of the pros we have much higher deterrent effect. And two, deterrence when combined with officers that have very good skills, it's able to reduce harassment regardless of its form if it's severe or mild.

Jennifer [00:44:09] I think this is also really interesting because it suggests that, you know, I was kind of saying earlier, you could just put a cop on a street corner and they can just stand there in uniform and deter bad behavior, but this suggests that they're doing more than that. Right. So they might not be arresting the person, but there's clearly a difference in what these police officers are doing with the ones that are taking mild harassment seriously and the ones that aren't. So maybe they're you know, they're intervening in some way or telling some guy to cut it out or something while they're there at the hotspot that is changing everyone's behavior and pushing people to behave better. Is that right? Is that the way to be thinking about this?

Sofia [00:44:45] So yes. And this comes to the really micro aspects of the difference. I think we haven't measure very well because you can be that they're patrolling in a corner and do nothing, but even how you position yourself, where you are looking, the way you are looking, we are always communicating with each other and it's the same with police officers. So it is likely and this is of course something we don't observe in our data, but the way officers communicate with civilians in a verbal or non-verbal way is likely to be different based on their underlying social norms relating to harassment.

Jennifer [00:45:25] Yeah, that's really interesting. Okay, so what are the policy implications of these results? What should policymakers and practitioners take away from all this?

Sofia [00:45:33] Yeah, so this is really exciting because after combining the lab results and combining what we've learned, we define how officers skills and norms correlate with the performance of "SHE Teams." I think the first policy implication is very clear.

Sofia [00:45:49] We need to equip officers in better ways if we are preparing the state actors in a better way to handle gender based violence offenses. So in this case, we're looking at sexual harassment, but if we're going to extrapolate to other types of offenses, we would have this very clear implication that, okay, they need technical skills, but there's also this component of norms and soft skills that officers can have better job performance. This is really the first policy implication.

Sofia [00:46:29] Then the second implication from this paper is that so police this is the tool we study and I think it is a really good first step, but sexual harassment, as we've seen in the paper and hopefully through this discussion in the podcast, it is a complex phenomenon and is very widespread and it's very high rate. So it's unlikely that we're going to address the problem just solely relying on the police and there is a lot that needs to be done in terms of other prevention, early prevention and changing mindsets of the society at large and same in schools or in communities things of this nature.

Jennifer [00:47:12] Have any other papers related to this topic come out since you all first started working on the study?

Sofia [00:47:17] No. So in this topic, really relating to safety and how it impacts women in harassment from the lens of police not much, but of course there are work ongoing in the context of Pakistan, for example, trying to establish this link between reducing women's constraints to mobility and how it impacts their job outcomes and constraints through mobility would also imply improving the safety in which women commute, for example, but to my knowledge, not more no.

Jennifer [00:47:54] And so what's the research frontier? What are the next big questions in this area that you and this dream team you're working with and other researchers will be thinking about going forward?

Sofia [00:48:05] So in this group, so we all very passionate about trying to understand police and the police in developing countries or lower income countries, because, one, there's not much understanding about how can we improve this institution because of course, as you mentioned, this is really difficult to forge collaborations as you know so well with the police.

Sofia [00:48:30] So we don't understand very well, even though we've seen papers in other contexts, that is very important for economic development, economic growth and integration of all the individuals within the society. So I think this would be the big one of the first the next steps and the second within the aspect of sexual harassment we really don't know nearly enough how to address it and we don't understand really well how women cope, how women make decisions and behavior and their behaviors around the topic of sexual harassment and how the how this impacts their lives. It's very micro-level, which I think is something we're going to keep on trying to do study in the future.

Jennifer [00:49:20] Lots more work to do. My guest today has been Sofia Amaral from the EPO Institute in Munich. Sofia, thank you so much for talking with me.

Sofia [00:49:28] Thank you.

Jennifer [00:49:35] You can find links to all the research we discussed today on our website probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts. To make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show and thanks also to our patrons, subscribers and other contributors. Probable Causation is produced by Doleac Initiatives, a 501(c)3 nonprofit, so all contributions are tax deductible. If you enjoy the podcast, please consider supporting us via Patreon or with a one time donation on our website. Please also consider leaving us a rating and review on Apple Podcasts. This helps others find the show, which we very much appreciate. Our sound engineer is Jon Keur with production assistance from Nefertari Elshiekh. Our music is by Werner and our logo was designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.

Jennifer DoleacRCT, Policing