If we could have session 3 mobile data for transport speakers come up, so this session will be led by Nicolas Peltier, who is the incoming global director for transport at the World Bank, and then if we could have Garrija, Rob, Alice, and, uh, Florence, your placards should be up there. We've squished one more chair in there for the next two sessions. All right, good, uh, good afternoon, uh, uh, everyone. Um, so my name is Nicola Peltierti Thiiberge. I'm, uh, I'm currently a practice manager in, uh, Latin America for, for the transport sector, uh, and in a, in a few days, I will, uh, join our global department. Um, so I'm very, uh, very, uh, thankful for, uh, the invitation and for, I'm very honored to, uh, moderate this, uh, this session on, uh, mobile data for, for transport. Uh, big data, as in, uh, as you've heard in other sectors has quite revolutionized our, our business in the, in the transport sector. We, uh, thanks to this new data which are much cheaper than what we used to do in the past with individual surveys, we have a much, much better understanding of mobility patterns, of mobility behaviors, and this is a great asset for when you work in. As a transport professional because you can understand better what, uh, how people move and what, what they need, and you can design uh uh mobility programs that are best fitting those those needs. No, it's also of great value to uh understand the uh the benefits of our, of our transport programs. For example, we routinely Evaluate what what are the benefits of mass transit investments on the job market in large cities. In particular we are now capable of measuring very easily what's the percentage of the jobs that are accessible within an hour of commute. Uh, in a given city, uh, and when we invest in a, in a mass transit system, in a metro, in a bus, uh, rapid bus transit system, it's, uh, we can measure those benefits on the, on the labor market, so, so it's, it's, it's certainly an area where we need to do more, um, where we've, uh, which has already changed our business, and I'm very excited to. To hear today about those uh those 4 examples of work that we've uh that we've done, we are very pleased to be joined by 4 distinguished researchers who and colleagues who will present some of the analytical work that they've done. Uh, to better understand mobility patterns, to prevent sexual harassment in public transportation, and also to assess how road safety has evolved, uh, during the COVID pandemics, uh, and, uh, with the associated mobility restriction policies. Um, let me say a few words about our, our, our speakers. So first, uh, it's my great pleasure to introduce, uh, Girja Borker, um, who is an economist in the World Bank's development Impact evaluation department where she leads the research on gender, and Girja will be presenting on measuring the incidence and reporting of violence against women in public spaces in Dar es Salaam. Then we'll hear from Robert Marty, who is a research analyst in DM. His work focuses on leveraging innovative data sources for impact evaluations of transport corridors and road safety. We will then have a presentation on private and public transport in Lagos, and we'll hear how mobile phones can help researchers understand mobility by Alice Dio, who is an economist in D. And last but not least, Florence Condeli will present her work on demand for safe spaces, avoiding harassment and stigma. Florence is a senior economist in D. So in the interest of time, we'll do all 4 presentations back to back and we'll keep the questions at the end. So if you're joining in person, uh, uh, please hold your questions until the end of the session. If you are joining remotely, you can share your questions in the chat at any time. And we'll be prioritizing questions from uh the live audience, uh, and I understand there are speakers will also remain uh at the end of the sessions um on Webex, so if you have any other burning questions, you will still have an opportunity to, uh, to ask them, no. So without further, further ado, let's start with Gerard's presentation. Gerard, the floor is, uh, the floor is yours. OK, so today I'll be talking about how we can use mobile technology to measure the incidence of violence, uh, that women face in public spaces and also improve the reporting that we see. So violence against women in public spaces is pervasive and it's a problem sort of across the world. If you look here, this is just anecdotal data or qualitative evidence that we know, uh, that these rates of harassment are very high. So a quick way to read this in India, for example, 79% of women have faced some form of harassment while moving around the cities, um, in, in their lifetime. So this sort of violence is also very unique in that it poses a unique risk to women. It's very highly frequent, uh, there's a large variation in types, so ranging from staring that makes a woman feel uncomfortable to a full blown assault, and also it's highly stigmatized and highly normalized. If you face it, you're stigmatized, but if you face it, you're also told, oh, it's just life. Um, there's also specifically in Dar es Salaam, high rates of violence that women face, so there's qualitative data that indicates women face groping, catcalls, inappropriate comments, and assault very commonly as they move around the city. So the extent of this problem is really unknown. Um, there's a big, uh, impact evaluation that was led by DM in Dar es Salaam, uh, related to the bus rapid transit. So we sort of, you know, just piggybacked on that and collected some data, uh, on 207 women, and what we found there was that 59. 90% of the women who were interviewed in Dar es Salaam have experienced some form of gender-based violence, uh, have faced some form of violence at least once in the past 6 months. Most of this violence is concentrated on women who are between the ages of 18 to 25, even the older women also face such harassment. Now if you compare that number of 59%, and I'll sort of keep coming back to this, in 2017, only 173 cases of harassment were recorded by police gender desks across the city. That's less than 1 case per 100 women in a year. Um, and these police gender deaths are spread out all over the city and they're supposed to be avenues for women to very easily report such type of harassment. Now, this sort of harassment imposes really high costs on women. I mean, first of all, it violates women's fundamental right to movement, to personal security, uh, to just feeling safe no matter where they go, but beyond that, this imposes huge psychological costs. Women take precautions to avoid such harassment and that avoidance behavior further imposes costs. So Fran is going to tell us about the travel costs that are imposed by such harassment and also I have work which shows that women are sort of losing out on higher education because of such harassment. Now the question is we are in, you know, we all care about policy and we want to solve these problems, but how do you solve a problem that is not understood? No one knows the scale of this problem, so poor data, weak reporting systems, no analytics, and lack of policy experimentation makes this even more challenging, um, so that is where this project. Comes in, we really are trying to develop a measurement tool which helps fill this big data gap that exists um in the system right now. So we are developing a GBV incident reporting app which allows women to record and I'm gonna get into a lot of detail um on the app but it records it allows women to record the harassment that they. Experience that they witness and also that they perceive because perceptions drive behavior so we want to record all of that and we also want to measure the gaps in reporting so sort of compare the data that we are getting to the administrative data that is already out there um we're going to be doing some policy experimentation on what can. Promote women to report such harassment, not just in the app, but also to the police gender desk, for example, and then we're going to go on to measure the effects of, Harassment and this reporting on women's mobility, their mental well-being, their self-efficacy. Um, so, uh, you know, the main goal of this project is to really push the frontier on measuring GBV. Uh, there's a lot of work that has already happened, um, in terms of, uh, you know, surveys that have been conducted, but we're really trying to measure this as in, in as granular way as possible, um, across space and time, um, and also really trying to make sure that the cost for cost. Two women of reporting are lowered um and we're building on the work that has already been done. We're not the first ones to develop an app, but we're really the first ones to build in policy experimentation in it and also take it all the way to respond to the needs of multiple stakeholders. So this app has been developed in collaboration with multiple stakeholders on the ground and is trying to build in all the features that are needed by them. So without further ado, let me show you, uh, Gianni, uh. Or on your way, uh, so this is an app which allows women to record the incidents of harassment that they face and also the perceptions of safety as I said, uh, it will include both incidents that are experienced as well as witnessed because we're really trying to get a sense of incidents and as I said, uh, you know, it's a stigmatized issues so a lot of times women feel comfortable in saying it happened to someone else, uh, when it's actually happening to them. Um, it also allows women to record all forms of harassment, so ranging from the whole staring to a full blown assault. Um, we will record the exact location and the mode of transport because we really want to, you know, as Nicola said, we want to bring it back to what can we do in public transport design and infrastructure design to solve this problem. Um, and, and finally, women are also going to be able to see sort of how much harassment they've recorded to really, you know, change begins at the individual knowing the scale sort of makes women, uh, speak up. Um, the way we're going to do this is we are focusing on the most at risk at the moment, so 18 to 25 year old women. Um, and, uh, these are women, you know, Sweta said it, we need smartphones for them to use the app, so we are limited by that. Um, they own their smartphone, and we are focusing on women who are enrolled in universities in Dar es Salaam for two reasons. First, cost, we want to recruit at limited. You know, fewer locations, um, second, uh, these women sort of have, uh, you know, they can use an app they'll record the data we expect, uh, and finally we are going to in the next phase of the project look at transition to the labor force, so we're picking courses which are sort of professional and where women are expected to enter the labor force. Uh, just to reiterate, um, uh, we are not asking the users to change how they travel in any way. They go on, go on about their life, and we're just asking them to record the harassment that they face when they're at the locations that they're anyways at. Um, and the experiment that we're doing to really measure like how women report, how much they report is that we have one, we have 1200 female students who are recruited. Um, and, uh, we will have 300 women who are paid and incentivized, uh, to use the app. So this is just simple check-ins. They literally just need to open the app and that's OK, but we just wanted to make sure that there's enough usage to record incidents. Uh, we have an unpaid group and this is the group that, uh, will receive no training, you know, this is where the scalability comes in. This group will tell us if we were to launch this app tomorrow. Put it on Android or Apple then how much usage would we see um and for them we will also be doing um uh sort of policy experimentation with behavioral nudges uh so if you tell women that what you do in the app makes other women feel safer then does that promote reporting so we'll be doing, um, uh, you know, sort of checking these scalable um uh interventions soft touch interventions as well. And then every time we're comparing how much they report in the app, so these are the people who are receiving the app to people who have no app but are using traditional ways like surveys of reporting, so that we can tell you, Administrative data tells us very little, surveys do a little bit better, but how well does the app do? So that's something that we need to figure out. Um, I want to show you some preliminary results from a pilot that we conducted with 30 women for 2 weeks. So this was the beta version of the app that we were testing. But even with this, I mean, it's, it's amazing what you see. So, these 30 women in 2 weeks have recorded 35 incidents, which translates to 903 incidents per year for these 30 women. So that's 2 incidents of harassment per month per woman. I told you I will say this again, but just look at the number that goes in a year to the police gender desks. So you know, this, this is a serious issue, a big, big problem when you think there's 5.8 million women in Dar es Salaam alone in this age group who are facing such incidents. Um, and the harassment that they face sort of ranges across the board, so you know, you've got staring, um, you've got catcalls, you've got a lot of unwanted touching and stroking going on, both that is amongst witnesses and victims, and that is witnessed. bulk of this is taking place when they're interacting with public transport. So 58% of incidents are when they're either waiting or traveling in public transport, and almost 90% of the incidents are when they're traveling, so when you include walking as well. Um, in Dar es Salaam, most of the incidents are taking place in Dala Dala, which are sort of also the most commonly used mode of transport, and this is so revealing. There's almost an even split between the perpetrators across service providers, which is bus conductors and bus drivers and passengers. This we can address this we can do something about very quickly, um, and, and that's important for policy. And finally. None of it is reported, so 71% incidents are not reported to anyone. The ones that are reported are reported to family, police draws, and that's completely in line with what we see. So, you know, with this limited data, we can tell that this is a major, major problem. Big public issue, public health issue, which is not understood, it imposes significant mental, physical, and economic costs on women. We are trying to use mobile technology to provide women with a tool and also us to understand what is going on, how much of a problem it is in a major city like Dar es Salaam, and also conduct experiments, sort of take the first step to understand what can promote women to record this, how can we solve it, um, by, you know, seeing more data on it. Um, and as I said, this is just phase one. The hope is that we will use this tool to extend the work to develop policies that can truly start to address this harassment that we see. Thank you. Thank you very much, uh, Gerard, fascinating, uh, story, and, uh, you know, we've, uh, we've heard similar stories, uh, in, uh, Latin America, in Africa, in AA, in Asia, everywhere, you know, it's, uh, and what is really nice in your, in your presentation is that there is a, a, a simple actionable, uh, uh, in terms of the service providers that we can do something about it, uh, at the policy level, no, so, uh, great, uh, great story. Uh, we will continue with, uh, Robert. Perfect, thanks so much. Um, so this is a project looking at the interaction between COVID-19 policies and road safety with a case study in Nairobi. Um, and it's a project in collaboration with a bunch of colleagues at the bank, um, and also with FLIR, which is an emergency ambulance emergency dispatch platform based in Kenya. Um, so first, give some context on road safety and the COVID policies in Kenya. So, Um, first, looking at, uh, Kenyan comparison to countries in Africa and also globally, Kenya has a bit above average, um, cases of road traffic fatalities compared to Africa and also globally as well. And in terms of COVID policies, uh, Kenya implemented, um, a number of the policies you saw, uh, globally as well, so there were policies to limit, um, Uh, social, sorry, uh. Policies to limit social mobility such as uh closing schools, banning um work from home orders, and also one of the big things that they did was implement a curfew from 7 p.m. to 5 a.m. and this was fairly strictly enforced, especially at the beginning when there was a large police presence out there. So pretty much, you know, after 7. 7 p.m. no one was really on the roads. So for our research question, we're looking at the impact of COVID-19 policies on road safety outcomes, and we leverage the policies as a natural experiment to look at, you know, what happens to road safety when you suddenly have a significant decrease in the number of people on the road and a decrease in mobility. And I think what you expect to see happen isn't entirely um clear. So you have lower mobility, fewer people on the road. I mean, just through less people on the road that could lead to less crashes because there's less road users, but also less people on the road could lead to higher speeds, which leads to more risky driving, which could lead to, you know, either more crashes or also more severe crashes. So we're not the first to look at this interaction between um COVID-19 policies and road safety. Um, so there's a few studies that look at this where the general finding is that it seems that there's, um, the policies resulted in a reduction of overall crashes just with less road users, but some areas are seeing an increase in fataler or more severe crashes. But most studies are focusing on more developed countries, where there tends to be um more data on crashes. So our big contribution is showing the potential first of using multiple data sources to overcome data limitations, especially in a bit more of a data poor setting. Um, and also looking at a developing country context where the impacts of COVID mobility policies could be different due to different mobility practices, um, in a lower income settings such as higher use of motorcycles and a lot more pedestrians, um, along the road. So in terms of crash data, we're gonna rely on 3 different sources and we rely on 3 different sources because each source is gonna have its own set of biases. So, bringing everything together, we can try like triangulate the impacts, the triangulate the impacts of these policies by looking at the common trends among these different data sources that kind of come with their own issues. So first, we digitize over 17,000 paper crash reports from the National Police Service. Um, so this is gonna include information such as characteristics and severity of the crash. Second, we partner with Flair, which is, uh, an emergency response dispatch platform based in Kenya. So we use their data on emergency dispatches to crashes. And third, we scrape tweets from Maker Root, which is a Twitter-based platform that's um over a million users that's based in Kenya where people commonly post about crashes in, in traffic. So we developed an algorithm to identify and geolocate crashes um from this kind of online crowdsourced source. And then we rely on a number of different data sources to get a congestion and speed of mobility, so we get congestion data from Waze, mobility data from Uber, and speed data both from Google Maps, um, and Uber as well. And then, so, to look at the causal impact of these policies on road safety outcomes, so our main method is gonna be a difference, a difference approach, uh, where we look at the difference both before and after the policies were implemented, so before and after March 16th when the first policies um were implemented, and then we compare crashes in 2019, so before COVID and then in 2020 after the policies were implemented. Um, and we're gonna use data both from 2 weeks to 2 months before and after policies, um, were implemented to look at, you know, the sensitivity of results to the bandwidth use, um, in both looking at the impact of crashes generally, we also look at the impact the policies had on specific types of crashes, looking at crashes involving a pedestrian, a motorcycle, and a matatu. So looking at trends in the mobility data, so first is the, the graph on the top is looking at trends in uh mobility data from Google, where, you know, right after the policies we see a fairly sudden and immediate and large drop in mobility. That slowly increases over time but never returns to pre-pandemic levels, and this mobility in turn causes a reduction in congestion, which is the graph on the bottom where that data is coming from ways where there again we see a fairly sudden and large drop in congestion that again slowly increases but. Never really returns anywhere close to pre-pandemic levels, um, even, you know, kind of months after the policies were implemented. And what's interest, so while we see an overall reduction in um congestion, you know, one interesting thing is with the 7 p.m. curfew in some locations we see an increase in congestion in these um locations. So with this in the roads in the red, we're looking at um congestion in the hour before the curfew, and the roads in the red are roads where you see actually an increase in congestion compared to pre-pandemic levels. So again, an overall reduction in congestion, but the, the Uh, lockdown specifically, um, did increase during some time periods, which is gonna interact with crashes as well. So looking at trends and crashes, so right after the policies were implemented, we see a sudden drop in um crashes occurring both when you look at data from Flare, the police to Twitter, um, but we'd see roughly after 3 months is that crashes are returning to roughly pre-pandemic levels when particularly when we look at the flare in police data, you don't see that so much in the Twitter data. We think that's because there's less reporting on Twitter because there's less people on the road to report on May rope. Um, and then here formalizing this in the difference in difference, um, so here looking at the difference of different coefficients across Twitter, please and flare again showing a reduction in crashes, um. And then when we look at the impact on specific types of crashes, the main consistent result that we see across police and flare is that there's actually an increase in crashes involving motorcycles, which is kind of in the middle of those graphs. So looking at trends across severity, um, you know, looking at the most severe crashes, kind of like crashes overall, we see a sudden drop in crashes, but again return to pre-pandemic levels. What's interesting is that these severe but not fatal crashes, um, again, there's somewhat of a drop, but they're increasing above pre-pandemic levels, um, kind of in these later months. And so focusing crashes by type, so looking at crashes involving Matatu or a minibus, uh, these return to pre-pandemic levels after a drop. Motorcycles, we never really see a drop. It just continues to increase and is getting larger, um, to pre-pandemic levels in the months after, and then crashes involving pedestrians. There's a reduction, but it is again returning to roughly pre-pandemic levels. And just focusing on this um result with Matatu is this increase that we'd see when there's no real drop. Um, so our thought is that this is due to just, we're seeing a lot more motorcycles on the road during COVID. So this table is showing registrations, vehicle registrations in Kenya in 2019 and 2020, where motorcycles were the only vehicle type that saw an increase in registrations in 2020 compared to 2019. Um. We think this is happening just because of a large expansion in delivery services which are largely being done through motorcycles in a in a place like Kenya. And then, so here we're looking at the how crashes change change over time. So the x axis is the hour of the day and the panel on the left is showing the distribution of crashes across hour where crashes tend to be a concentration around rush hour and the middle is showing the 7 weeks after crashes and what's interesting to see is especially in the police data, um, the distribution is super focused on the time, um, around. When the curfew, what was in place. So again, this is time, again, some roads saw more congestion. Post policies compared to pre-policies. Um, so again, this curfew and people rushing to get home may have contributed to more crashes during this time. And then looking at the changing crashes across space, the big finding here is that we see a big reduction in mobility that we're noticing from the waste congestion data and also a big reduction in crashes and specifically in the central business district. So there's, you know, less residential areas where people aren't going to work, um, we're seeing kind of less people there and also less crashes. So, bringing everything together, this is a table where we just take all the indicators that we use that we're looking at the percent change, um, both before the curfews or before the policies and after um they were implemented. And the big takeaway here is that as mobility decreases, we don't see proportional decrease in crashes. So looking at a couple of the crash sources, we see about a 29 to 34% in the reduction of crashes, while we see a much larger reduction in mobility, and we saw this from the figures before where we See congestion and mobility from Google, significantly reducing, and they never really returned to pre-pandemic pre-pandemic levels while it only takes about 3 months for crashes to return to pre-pandemic levels. So this suggests that for vehicles on the road, there's a greater likelihood of crashes. Um, thank you so much. Thank you very much. Uh, we had a very interesting presentation as well, uh, uh, with, uh, unexpected, uh, results of, um, the, uh, the COVID, uh, situation and the, uh, and the lockdown. No, I mean, we all saluted the, uh, The the reduced congestion and the uh the cleaner air, but there were some other consequences, no, and also with some interesting policy implications, uh, what do we do about it? Can we adjust enforcement policies can perhaps target it more particularly at the two wheelers, no, which there have been also a change of mobility behaviors, um, we'll now switch to Alice. Thank you. Hi, hi everyone, it's a pleasure to be here. So this is kind of preliminary overview of what we intend to do with mobile phone data because we just received the mobile phone data. So this is a project that is taking place in Lagos with. My colleague, uh, Dan Dongeon, which I feel, uh, like is, uh, very, uh, everywhere in this, uh, uh, today, um, and, uh, Gita Nakpan from both from Brown University and, uh, Nick Tivanides from, uh, UC Berkeley, um, the idea of this project is to try to have, um, An understanding of both uh public transit in a rapidly growing city. Lagos is like rapidly growing and on its way of being the biggest city in sub-Saharan Africa by 2050 and at the same time also understand what is the dynamic of um informal transit because a lot of countries will not be able to afford or cities will not be able to afford uh those public buses. And in Lagos we have both a very vibrant informal transport sector but also a huge uh initiative on public um uh on adding new bus lines in the city. So we're taking advantage of that initiative to study both the impact of adding more buses on overall mobility in a city that is very congested but also we want to study the dynamic of informal transit itself in the hope of. Helping other cities, understanding how to regulate some of the externalities that uh usually the informal transport sector has. Uh, in addition, we also want to add a gender component in the commuter survey and I will talk about it in a second. So in terms of uh background and objective we try to use data to inform and provide uh uh evidence for uh better policy making. It's a collaboration between um LATA, so that's the transit agency in Lagos, World Bank, Brown University and UC Berkeley we're working very closely with the Lagos state Authority uh for this so they they actually pressing us to go ahead with. This with the commuter survey and so we're interested in understanding first the impact of that bus reform intervention that is adding 800 buses on the roads in Lagos and we're interested in understanding the impact on commuters does that like facilitate transit, shorten the trips uh what are exactly the the the demand for that uh type of transit and then. Um, second, we're also interested in the response of the uh, informal sector. For those of you who work in the transport sector in urban context, you know, all regulating taxis and informal transits, um, is difficult, right? And so we want to understand what they are responsive to, what is the evaluation of different routes, is there a way that we can have a. Better transport sector by looking at it like holistically in some ways that the informal transit might still have some space somewhere um and by informal you still have a certain amount of regulation because you have unions of transport worker association with motor parks that are set by Lago state so it's not completely anarchic right? and. Of course we want to know if it reduces congestion because Lagos, Lagos Island, which is down here, is really the the business district of Lagos and everybody's trying to get there in the morning, which is not the easiest thing to achieve. I've been stuck on that bridge for a long time, sometimes, so. So we want to understand how much the commuter do value the new system and some commuters will always prefer the cheaper informal transit, some commuter value the amenities that's the Uh, formal transit brings right, um, in terms of wait time reliability, but also there is AC in those buses and that is very convenient in a city like Lagos. Others the then for those are the informal mini buses, uh, sector respond? Do they increase prices or decrease prices? Do they change the way the supply. Uh, minibuses, do they focus on areas that are not as served with by the formal transits and, um, and then the next question is from a design perspective, what are the design that in terms of both routes served but also fares and wait time that would maximize the benefits, the total welfare of the population, so. Um, we have research questions that have disappeared apparently, uh, so the first research question was how do congestion react to that, uh, to that, um, change in public services. The second research question we want to answer is how do commuter react and for that we have both surveys and telco data and the third question was. Uh, around the behavior of informal, uh, operators and how do they, um, react in increasing competition so and for that we also use a mix of survey and observation data so literally people sitting at bus stop or using apps to look at what is the pattern of the informal sector because nobody really knows. I mean there is an informal knowledge of that but we need to be able to map this and so that's the. Planned for the summer, a huge census of the informal transit network. So what is telco data um telling us here? So you have two types of telco data and and Zera has has touched upon that already. So I'm not gonna um linger too much, but you can use either the CDLs without usually the phone call records and so you need to have the phone switched on. The problem is. Is in Lagos a lot of people uh don't want to use the data so they will just like not connect you still have the home registry that you can use with that so in this case in in places where data are expensive for in and the mobile phone plans might be a huge burden then you still have that option of using this so what we requested. Access to um it's the anonymized home location register so account identifier, the time stamp and the tower again like what is the nearby tower so we know where people are in the morning, where do they go in the afternoon, and when do they go back to, right? uh we need to of course know where the towers are, otherwise it's difficult to map um. And we can anonymize that as well and then we wanted to have access to past home location because of course that is part of what is costly in this data is just like it gets big really fast, especially if we wanna go back in time so often telco operator have that but it's kind of storage cost or maintenance cost or or or put it on the server cost, um. And so what do we want to do with that? So we want to measure um traffic, uh, speeds and um travel demand patterns uh along the roads that are both the ones that are treated and the ones that are untreated by um the BRI. So by treated we mean that the one where a bus line opens, what does it change? The congestion increase if the mini buses are still on it and at the same time you have public buses. It can get worse, right? Or is this like kind of uh uh uh a net positive? What we want, uh, uh, as well is try to um understand what is the travel time per mode. So on the places where you don't have um public buses, what can you do uh to recover that so especially for the past data, right? We want to understand what was the kind of typical pattern of mobility before, before the introduction of the buses. And we want to track all traffic and commuting patterns changed uh in response to the BRI not only on the kind of short term side right you can have like suddenly people switch from one mode to the other which is great for the transit agency but what would be even better if like more people were um able to access jobs thanks to the new offer of public transit. So, uh, the transport authority is very keen on us doing this work because for them they want to know where are the traffic bottlenecks a bit more, uh, deeply they already have an idea they have like people that, uh, analyze this data but basically we can do that at a scale that, um, they might not be able to do um we can also measure the demand in a finer, more granular way and, and, um. And help them optimize schedules so um you do have some kind of uh ITS system and and embedded system in the buses, but scheduling uh in Lagos is really hard because congestion happens sometimes because you have a rain or things like that so it's good to understand what will be the adverse event and suggest improvements there, um. And then uh we want to impact of course the uh the study the impact of the BRI and um and we want to put all of that together so the demand analysis, the bottleneck analysis and the. Behavior of the minibuses analysis to create some kind of model that would help like optimize this network right and think like OK if you price this there on that route you can catch that demand and then it will attract other people. So, um, what we're doing here is um comparing uh treated and controlled routes and of course for congestion, it would be city wise, right? So you can see um the, the routes that are treated in brown and so they increase of course over time and then you have spillover routes that are in the catchment. Area that we want to understand as well like because all the network is about connectivity so looking just as one link doesn't make that much sense and then we have control route that were uh untreated but part of the plan and of course the decrease uh this number decreases over time so an example of prior work done by one of my co-author um. It's an example of using the CDRs to see like where people live in which type of neighborhood. So you have a big migration in Jordan it's it's taking place in Amman and then they wanted to understand that what is this influx of refugees doing a bit for people do the patterns of mobility change because you have more people changing that and so you can have that at a very, very granular um. Uh, level and again, uh, we're looking at the impact of, uh, COVID here in terms of the risk mitigation, um, and this is something like Zveta touched upon and that's why we took so long to get access to this data because, um, we had several hoops in order to access it right and to ease the concern you have way of, of, of, um. Of uh getting into agreement and uh data infrastructure that is very safe, uh, and, and make sure that the both the telco but also the regulator are OK with it so we decided to leave the uh data stored, uh, in Nigeria so it's like on secure server as uh MTN is the provider that we have an agreement with um who has access is the Transport Authority through a secure VPN. And we cannot copy the data through the collection. What we can do is run queries and not like uh have any uh results that reveal behavior of individuals. So basically we can just produce aggregated results and we had like nondisclosure agreements. Thank you very much. Thank you very much, Alice. Uh, another, uh, great presentation, and we have a particular interest in the Lagos BRT because it was financed by the World Bank and it was the first BRT in West Africa. Now we have a similar systems in Dakar and Abidjan, but this was the Lagos one was the, was the first one. No, and, and for those of us who worked on this kind of project, the political economy of those uh. Investments is difficult and in large part because the informal transporters are usually pushing back on having more competition and so having this kind of understanding of the possible impact and finding a solution that can benefit that can be win-win to to everyone is is very, very useful. So thank you, thank you very much, Alice, and we'll finish with Florence. Going back to um sexual harassment in public transportation. Thank you very much. I'm very happy to be here today. Nice to have an in-person conference again at the World Bank. Um, so today I'm gonna very much follow in Girija's track, uh, to tell you about the cost of public, uh, sexual harassment. So this is a joint work with Arianna Legovini here present. It's part of a larger set of projects that are ongoing. Doing as part of our Brazil portfolio, so actually a lot of interest in infrastructure work, uh, for those in the room who are working on that, happy to, we're all happy to chat about it, uh, and it's joint work with Kate Vibni who's at Duke University who was, uh, a very integral part of this project from the start. So the motivation again, I'm not gonna over motivate this topic. Everyone's aware in this room. Uh, that you know, sexual harassment, particularly in public transit is a really big issue, right? The question is exactly as Gerri Jaff framed it, and this project is gonna be very much in that vein of saying, OK, so we know about prevalence, right? So we have surveys of, uh, users of public transports on the platforms typically that kind of ask women, OK, how, uh, you know, have you ever been harassed? But this and like this type of statistics is gonna sort of really mask a lot of potential heterogeneity and exposure to harassment, uh, as well as kind of the intensive margin of that behavior, and I think Girija really spoke to that. It's like, OK, maybe it's 35% of the time that a woman in her lifetime gets harassed, but how often when she makes the choice to commute daily, right? So this is very much, you know, the spirit of this project, um, so we have some evidence, and I'm not gonna over motivate this, some great work by Girija. Uh, that you know there's gonna be women who are going to remove themselves from top schools, uh, that they could go to uh to avoid certain unsafe routes on their on their travel to school. Uh, there's gonna be a lot of women who just decide to stay home and not go, not participate in market activities because going to uh to work would be too costly for them in terms of harassment, uh, but a lot of this evidence again is going to be very extensive margin behavior so I go or I don't go uh and. A lot of it is gonna be done in very conservative, uh, high crime context, right? And so what we know less about is in those contexts such as Rio de Janeiro in our case, uh, what happens to this woman, right? If you have been to Rio, it's not a city. I mean, sure, there is crime and stuff, but there are so many women participating in the economy, you know, this is not, uh, a place where you would think, oh, there's stigma against women moving around, you know, and so the findings from this study for me were very surprising, uh, in that context. So one policy response to this harassment in the public place and this is something you've all been familiar with I think London uh has put in place a women reserve safe space on the, on the tube for example um this, this type of policies of creating a women only space in public transit have been pretty uh used, right? So the question is, and this is interesting and I think the the finding. From the studies are going to speak to that. They create a lot of very extreme responses, right, in the public. So some people will say, yes, of course, women deserve a place where they can be safe, right? And then others will say, no, this is terrible. You're creating a stigma against women to travel in a mixed, you know, environment. And so we're kind of, you know, blaming the victims for behaviors from, you know, a certain group of people. So this is a very, you know, when we started this project at the bank, we had many, many consultations with our operational colleagues and really the sense was so high that they hated this policy, right? I want to be honest. So this is the baseline with which we started this. So what did we do? So originally when we, and I'm just gonna tell you the story in this 10 minutes of how we did this project, I think it's a bit more interesting than eventually the findings which are striking, uh, but it's how this came about, right? So you enter this space and you think exactly like Grija, like how bad is this? And our first thinking was it can't be so bad. This is real, but you know, the government really wanted to work on this, so we thought, OK. You know, maybe this is not so bad. Maybe if we start interviewing women uh on the platform daily and getting, you know, tracking them and stuff, maybe we're not going to see so much harassment. Maybe we're not going to have statistical power to detect this. So we started to think, OK, one thing we can do is use this reserved space and find a payment like a pricing mechanism. Mechanism for the reserved space instead. And if we find that women are willing to let go of some money to be in that space, through revealed preference, we'll be able to say they feel probably there's some amenity value to being in that space, right? And then we can get some data and back out that maybe this is the cost. So there is harassment, we can back out some stats. So this was our first, uh, our first approach. And this is how we set up what we're, what I'm gonna present. Then in the next step, what we thought about is, OK, so once we have this data, and I'll show you how we generate the data and the type of data we have, we thought about, OK, it's interesting that people really feel strongly about this policy and it's interesting that perhaps some women choose to go into this reserved space, not because necessarily they are Afraid of crime, but maybe they're afraid of bearing the stigma of not riding this car and or if they're harassed in the reserve, in the non-reserved space to not be able to tell their friends and family about it because they'll feel some, you know, greater shame from having been in the mixed car and exposed themselves. So we're gonna tell you about some data that's going to speak to that issue. OK, so what's the reserve space, uh, in the case of Rio de Janeiro this has been put in place in 2006. We're not gonna study a before after forget that we're just gonna do something else for identification strategy. Uh, this is, you know, uh, reserved space it only operates during rush hour, so the rest. Of the time that pink car, it's, we call it the pink car, uh is uh mixed. So it's only between 6 and 9 and 5 and 8 that this is gonna be a pink car. So, our study is gonna only study uh people who are gonna travel during those times and when they have the choice. OK, so the data and experiment where we started with, so the first element of, of data that we generated was crowd-sourcing preferences and experiences from riders using an app. This is not an app we designed, we used uh premise data as our provider of the app. Uh they send people to do tasks around, get prices, do stuff like that. In our, uh, in our task setting we asked these female participants, 357 of them, uh, to, to tell us about their commuting experiences. We did not phrase it as anything to do with harassment. Uh, we just said we wanna know about how you feel when you travel in, uh, in this case the Supervilla. We paid the participants the equivalent of the fare, uh, so that it would be a cost neutral activity for them, so we were not incentivizing them particularly to take the train if they didn't need to, uh, but you know we reimburse them that cost and in total from these 357 riders we're going to capture about 22,000 train rides. And I'm gonna tell you what we do with that. So basically we ask them to ride uh the the subway and then tell us about what's happening, you know, uh, is the AC working? Was it crowded? Uh, were people loud? What's going on? This is gonna serve two purposes. First, they're not gonna think about harassment, uh, when they're riding. Second, we're gonna know about, you know, the local conditions. We're gonna be able to say a little bit about what happens of course this data, the problem with this is. This is part of the choice, right? So, if I choose to, if I choose to locate myself in a certain space, this may be uh in response to certain amenities that are in that space that day. So we're not gonna use this data on uh amenity value for anything. Instead, we're gonna also send uh a team, sorry, I'm gonna skip forward a little bit. We're also gonna send some men. Uh, to the platform in a different, in a different ask and ask them about how is it going on the platform today? How many men are in the reserve space? Is it crowded? Uh, how many, you know, how many men are in the non-reserved space? What are the local conditions? And so this is gonna give us some data on what's happening in all spaces during the rush hours throughout our study period that's not gonna be contaminated by the choice of our individual riders, OK? So we're gonna use that. OK, so then what we do with the app is that we price um the rides in different spaces differently and we do that slightly randomly so that there's no, you know, anchoring of expectation, all these things. Um, so sometimes you're offered a zero opportunity cost to ride in the reserved space, meaning the price is equalized. You can just choose in the app which place you're gonna, you're gonna be on your ride. You just say, oh, I'm gonna ride today, up reserve space, I go, you bear no cost for it. And sometimes there's gonna be a. Positive opportunity cost meaning it's gonna cost you, you're gonna get paid less if you ride uh in the reserve space and we're gonna vary the pricing and that's gonna give us a measure of willingness to pay to be uh sitting in I mean not to be standing in the in the reserve space and then in a third phase we're just gonna randomly propose that our riders ride in these different spaces. And then we're going to ask them in that last phase, when we're done with the experiment, we feel comfortable asking them, OK, what happened to you today on the commute? Were you touched, where you talked to? Did you experience any form of harassment on the commute? The reason why we did it at the end of the experiment was to avoid exactly kind of this feeling that this experiment was gonna be about harassment, but in that last phase we felt like first for IRB reason it was gonna be OK to warn them that this was gonna happen since we're randomizing them in two different spaces and uh and, and you know, kind of at this point it's clear what's going on in the experiment. Now, to back up the estimate, an estimate for the cost of harassment, obviously, we're gonna combine these two measures. So we're gonna be able to, and I'm gonna show you some numbers after eliciting uh revealed preferences to pay for the uh for the reserved space, we're gonna use the randomization across space to tell you by how much those harassment vary across spaces, right? And then using these two numbers, we're gonna tell you how much harassment costs at the intensive margin. There's a lot of concerns about ethics. I'm not gonna go over it in 2 minutes, but essentially we did a lot, uh, to ensure that the ethics were respected, that, you know, our riders first at baseline everyone had ridden in both spaces, so it's not like people in our experiment didn't know what was, you know, what was going on in any of these spaces. They had ridden everywhere during some baseline rides that we that we paid them for, um, and then we directed them to local resources in case of in case of reported harassment. OK, so what are the results? Very briefly, uh, we find that about 25% of our sample is gonna be willing to give up quite a bit, so kind of 20 cents. This is not a large amount if you think about it, but as a share of the payment, it's actually pretty significant. So we didn't want to give very high powered incentives for this, for. Ethical concerns, right? Uh, but so you're gonna find that about 26% of women are willing to give up, uh, this higher premium and it's pretty flat like the, the demand doesn't vary a lot by premium which kind of indicates that we were probably really, you know, getting a lower bound, uh, estimate, uh, of this. Then what we find which is uh sort of What we would expect is that the presence of men as measured by the people on the platform is going to uh so the low presence of men in the reserved space is going to increase this willingness to pay by a lot and at times when the presence of men is equalized between reserved and mixed cars we're gonna find no, no, you know, residual demand for the car and that's sort of uh that's going to highlight the mechanism that we have in mind which is sexual harassment. So when we randomize. These women across spaces we find that the incidence of harassment is going to be halved when they arrive uh in uh in the reserve space and the numbers we're going to find are a bit lower than those uh highlighted by by Grija in uh in Dar es Salaam but not that much lower. I mean this is once a month uh is the uh incidence of harassment that we're going to find in this uh in this mixed space. So this is very high, um, and we're gonna, you know, in the paper we study alternative mechanism what if. It's not just harassment and we're gonna rule them out. Um, so this is gonna tell us that basically for incident these women are willing, I mean this incident kind of cost this woman about $1.50. Uh, this is a really low bound estimate, but even at this low bound estimate, you know, the associated wage penalty would kind of shrink the labor market supply from these women by 0.5 to 1%. So this is like a pretty sales tax result, uh, from, uh, from this, uh, these results. OK, I'm not gonna have time to talk about stigma because I'm out of time completely, uh, but I just wanted to share with you that in our stated attitude survey that we do on the platform, uh, with, uh, with different riders we find that, uh, people are really willing to state to us during a survey that they think that women who choose to choose to ride in the public space. Are more sexually open and that those that are, you know, riding in the mixed space are kind of looking for trouble, so to speak and when we correlate this to the demand for cars, so women for whom this is the prevailing norm, women who tell us they think other people think this way, a second order belief, are going to be 80%. And more likely to actually ride or to tell us that they ride in the reserved space. So we think there's really something there that's not just protecting yourself from uh from harassment, but also protecting yourself from the stigma that's associated with riding the mixed space and getting harassed. OK, I'm going to stop here and then we did IAT but we're not going to talk about this today. Thank you very much, Florence. A really nice compliment to Guir Jar's presentation. Uh, we have actually a budget support operation going to the board in a couple of weeks that includes a policy in Rio de Janeiro to to fight sexual harassment in the public transport. So I hope that the situation will, will improve, but there are a lot of debates on this safe space. No, it's not the first time. Uh, that I hear that, it's quite passionate, uh, the experience of Mexico also has been, uh, quite discussed, so, uh, uh, but I, I can reassure you that, uh, in the transport family, uh, now it's, it's basically routine to, uh, when we start preparing an urban mobility program to, um. To, uh, to streamline this, uh, this topic in the, in, in project design. Oh, and actually I was in Chile, uh, a couple of weeks ago and the, the first thing that the, the Minister of Transport was asking is can you help us to make transport more, uh, more inclusive. So I think that, uh, mindsets are changing on this, and that's great, no. So, uh, Um, thank you very much. We can come to the end of the 4, presentation. I don't know if we have a few minutes for we can take a couple of questions. Perhaps we are already over time, maybe 2 or 3, or yeah. Anyone, Um, thank you very much for your presentations. Uh, this question for, um, uh, Grisa. Uh, the app does consider, um, the change in behavior of like different women. Like, for example, some may actually change the routes or even the modes as they have been, you know, exposed to harassment, and I think Florence mentioned a little bit of that. And maybe, you know, we may be missing some of the more, uh, like dangerous routes because simply women are not riding those routes anymore. Um, the second is, I wonder if you know, if the research in general, and this is for both for Chiria and Florence, uh, researching harassment should also be, um, should also have a component of like kind of like a support system for the women that are being surveyed because in some cases, you know, um, harassment and reporting harassment could be traumatizing for some of them. I'm just wondering if that will be a component of your research. And the last question is related to, you know, um, if some of these, if, if in your research you have seen some of the root causes of this, right? Because we're surveying like the women who are being harassed and are the victims. What, what about the perpetrators, right? And, I think Florence, uh, discussed a little bit about that, you know, what would be the policy making like moving forward to reduce this kind of occurrences. Thank you. Thank you. Let's perhaps uh take that, that question, uh, who would like to. To answer I Yeah, I think Grija's gonna have more to say than me on this, uh, but, um, what I thought was fascinating maybe to, to also take a step back is that there's a lot of heterogeneity in our case in who gets harassed, and perhaps this speaks a little bit to Gereja's point on the stigma is such that people. And you know people have sort of made this part of their lives. Uh, so we saw in our sample for example that 50% of the harassment is concentrated on about a third of women. So it's also quite interesting, um, you know, to think about what's accepted, how these perpetrators are kind of, you know, sort of. It's OK, it's normal. They're just men being men, you know, whatever. Uh, so I think to your point, I think some of the, the actions that people are taking on, on IPV and harassment is, you know, cognitive, uh, based therapy type of things and, uh, I don't know. I, I think Girijack can say a bit more about that, but in terms of support to, to people who report. Violence and this is something we have across our gender portfolio um is really like indicating resources where people can go to uh that are local, that are easy to access, not costly and just kind of a credible uh way of addressing the issue and all our research goes through IRBs uh so we're like trying to, you know, follow what's uh advised. So I just quickly to add, so I, I think France France actually touched upon most of the things, uh, but the change of routes, so you know how I mentioned that we're actually showing the women an activity log, like, you know, they, they will be able to see the harassment. A lot of women, you know, they say when we do qualitative, when we did focus groups that they just face it so often that they don't know the scale themselves. So we do expect that that will have a role to play and we will measure that as part of, for example, the, the Tanzania project. Um, in terms of support system, yes, IRB, we have this even inbuilt in the app so that they know who to call. In fact, we've tried to link the closest at that location, uh, help desk so that they can get support maybe quickly. Um, in terms of root causes, so there's very little that has been done, uh, in terms of, you know, perpetrators. Uh, there has been more that is done in environmental factors or infrastructure design, which contributes to women feeling more unsafe and contributes to harassment incidents. Um, you're right, more needs to be, uh, uh, done on the perpetrator side and, and, um, you know, hopefully, sort of with this work, uh, we will be able to. Thank you. Maybe one more question to have oh yes, there is one in the, in the back that would be the last one. Um, I guess this is mainly a question for Florence. I came in a little late, I'm sure Girria will have thoughts too. So you talked about the dimension that when uh, when there are fewer men, their willingness to, When there are fewer men, the trade-off of sticking to the reserved spot changes. You don't mind giving up your space in the reserved spot. I was thinking of the reverse, right? Like the root issue here is that public spaces are dominated by men, and there isn't a critical threshold of women that you see around. So it's like, do you see it in your data, or do you know of work in the reverse direction that is that a critical threshold of women in these mixed cars, um, and once that's crossed, then women feel safe and men also kind of hesitate to do something out of order. Um, yeah, I don't know. That was just the general direction of the question. No, that's super interesting. So, on the compliance to the rule, I mean, essentially we see pretty bad compliance to the rule in our data, so we don't really see any time of the day when there are really zero men uh in the, in the pink card which is something to highlight. The fact that we see this again very quick presentation, but the fact that we see this reduction in harassment, uh, as a result of being in that uh reserved space was surprising to us because we. Road there and we saw it and we're like there's no way this is working at all. So, uh, so that was our, uh, you know, biased opinion starting the study. Um, having said that, yeah, I agree there may be some tipping points, uh, both ways, right? There are too many women so nobody enters. I don't think our data can speak to that with the granularity we have, but, uh, and, and the transit data we have from the, the transport authority is only going to look at density of passenger. It's not going to look at kind of of course who's sorting themselves where uh that's the more granular data that we have but doesn't answer your question. So I think it's a super interesting, you know, what drives compliance uh and how to, we try to do a compliance pilot where we had uh agents from the from the the operators kind of be on the platform and really try to enforce. It didn't really work well. Uh, we got kind of a no effect on that so. Thank you so much to the panel. Thank you, Nicola. I really appreciate all the speakers and we'll take 10 minutes for uh for a coffee break and some treats outside. Thank you so much.

showAllTranscripts

duration

PT1H3M52S

scene7File

worldbank/Measuring Dev. 2022 Session 3

scene7Domain

https://worldbank.scene7.com/

scene7FileAvs

worldbank/Measuring Dev. 2022 Session 3-AVS

title

Measuring Dev 2022 Session 3

description

Measuring Dev 2022 Session 3

showTimestampAndTranscript

yes

sec-spacing

section-none

sec-spacing-bottom

section-none

sec-swoosh

sec-swoosh-type16

style

default, transparent-white-components, heading-center, add-style, bg-primary-blue-70-gradient, bg-dynamic-height

col-xs-12

col-sm-12

col-md-3

col-lg-3

col-xs-12

col-sm-12

col-md-7

col-lg-7

add-style
lp-body-content

Measuring Development 2022: The Role of Mobile Data in Global Development Research — Session 3: Mobile data for transport

lp-heading-top-medium

lp-heading-bottom-medium

Measuring Development 2022

https://www.worldbank.org/en/events/2022/05/11/measuring-development-2022-the-role-of-mobile-data-in-global-development-research

Opening Remarks

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-introduction-and-opening-remarks

Session 1

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-session-1-recording

Session 2A

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-session-2a-recording

Session 2B

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-mobile-data-and-covid-19

State of play lecture

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-lecture-recording

Session 3

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-mobile-data-for-transport

Session 4

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-methods-for-mobile-data

Lightning talks

https://www.worldbank.org/en/news/video/2022/05/20/measuring-development-2022-lightning-talks

col-xs-12

col-sm-12

col-md-2

col-lg-2

sec-spacing

section-none

sec-spacing-bottom

section-padding-between

style

default, video-transcript

everything

https://webapi.worldbank.org/aemsite/everything/search

true

{"facets":[],"orderby":"contentDate desc","select":"*","searchFields":"*","search":"*","filter":"contentType eq 'Video'","top":20,"skip":0}

See Rest of Series

yes

sec-spacing

section-none

sec-spacing-bottom

section-padding-between

style

default, bg-neutrals-15-gradient, remove-border, lp-mini-card-primary

Measuring Development 2022: Mobile Data for Transport

RELATED

See Rest of Series