 I just have a question about the empirical, do you use these kind of buffers to try to assess the fact, is it possible that you have some kind of intersection between buffers that you have some kind of contamination between one area and another area? How do you control that? In that moment we don't control on that, but we can check in this point it's important. I'm sorry, can you show us the slide again? Yeah, this one, the first one? Yeah. Yeah. In the left hand, I think there seems to be a bit of an appropriate trend. Yeah. So perhaps we will be nice and check something like the roughest matter for honest differences of appropriate trends, but we get a more green estimation. Oh, okay. Thank you. Yeah. Yes. We know that we have some differences before the policy implementation, but the trend is clearly that there is a shamp after the policy implementation that increases the total number of university students in those localities. But yes, we have to check about these differences before. Can you show your results for completion? Yeah. How do you interpret this decrease? This decrease, first of all, could be a composition effect of students that can enroll at the university in those localities. We are more students. Yeah. And then they complete that. Yes. But then you don't find this for first generation? No, for first generation, no. We have the opposite effect in this case. So what's your story? Yes. At first, the thing that we discussed is that there are worse students that enter at the university when this campus opens, that maybe they don't study if they don't have the campuses near their localities. But maybe the students that they don't have the opportunity to come to the capital to study, and they have lower costs to study, maybe they are. If it's for me that there are like two types of students that gain access to university, right? The first generation ones, where the ones that work good, but they don't have access, and then the extension, they gain access, and then there's another type of students, maybe, they were hindering them, and they didn't have access to the university, because they don't know the capital. And then maybe they moved to other cities to... Yes, they moved, or maybe... Yes, they moved, or they have the campus near their house, their home. Yeah, but this... Maybe they enrolled... For the first generation, most of them, they had access before, right? So the new ones that are gaining access here, they are worse in terms of... Yes, on average, yes. But when we compare, but when we analyze what happened with these first generation students, we observe that is increased. So maybe there are more... I think that there are two different stories. These two guys tell me different things about two different groups that are gaining access. I think it would be interesting to explore who are the people, they are like this group of these other groups, and there are very different results for them. And then maybe if you exploit them, a better idea of what's going on in the labor market, because then your labor market results, they seem very noisy, and you don't see anything there. But maybe if you focus on these different groups, you're going to... Yeah, it's a good idea. Have a clear story. Thank you. I should be worried about this, because you look at total enrollment for high school students, right? Students, they were attending high school in that region. Yeah, that's local. I'm worried that students are sorting in high school, they're moving to those regions in high school. Yeah, so we don't control of that, but it's important to discuss it in the paper. But because the information that we have is the locality where the people do high school. We don't know where the people are born. If you have information about migration for high school, maybe if you tell me old people that you usually don't move for high school, that's fine, but then it would be nice to do that. Okay, great. Thank you. What's the definition of first generation student? Like first generation student is a student that... It's the first time you go through higher education. No, no. It is the one that enrolled at the university, and their parents did not enroll at the university or tertiary educational level. So you have 80% of first generation students enrolled at the university? Yeah, on the maps with the... Yeah, it's a high percentage of first generation students at the university. So the students who enrolled at the university... Yeah. Their parents don't have higher education. They go better than students that... With their completion, yeah. You have 80% of first generation students, right? Yeah. And then you have a positive effect for your probability of graduation. For them, yeah. A negative effect for the general population. But then you have 80% of first generation. So you must have a really negative effect for the other group. Uh-huh. I think so... Yeah, it's the thing we find in the... Okay. Yeah. How are you fostering your second generation? Is that the geographical reason? At the locality level. How many localities? 140 localities. Thank you. Yeah. So it would be nice to have a normative statement on whether on average it's better off for everyone to have... So if whether... Like having students that enroll and do not complete whether that crowds out resources for other students... If it was more selective, maybe there would be higher quality education for fewer students. That's one thing. And the other one is the graph on your employment outcomes. Is it conditional on finishing high school? No. It's not conditional. It's for the average population between 21 years old and 40 years old. And on the labor market outcomes, we use the national household service. On the other outcomes, we use the administrative data. So there is a big difference between the sample we use here. High school. We can do it. We didn't do it. Thank you very much. Thank you. I don't remember about the program. Is there like a specific year that you can apply? Like just like finish the course or you can apply in the first year of the college? You need to have at least 20% of the courses. And no more than 80%. Because my doubt is like there's kind of worst selection. Like the worst students that they don't have like a job, an internship. For example, they have like an opportunity to cost lower. And they decide, I'm trying to do this, to apply to this program because I don't have an internship here. So I go there. It's kind of hope to have a better position in the job market when I come back. And the guys that are the best students, they are ready to have an internship. But I'm not losing my money. My internship here was you to do this kind of a job. Yeah, it's a really strong assumption, right? That's what I'm making. Like really strong because remember that those are students from federal universities. So they are on average better than the average students in Brazil. So yeah, I think it's a very strong assumption. I'm not sure if this... The comment that is supposed to be on this paper, take this opportunity. I think it's a very good job, as well as a good paper. Part of the effect, and also the effort of some of the makers while you are visiting Brazil. Part is, of course, a bit specific. Maybe if your channel is correct because it was Amanda Akimakura. Yeah, perhaps I was wrong. But the comment I'm going to make more generally is that I can't get this discussion with Shibiko. Sometimes in economics, we tend to be very precise about sort of small analytical exercises. And then we have difficulty interacting with the bigger questions. So for a second, if I try to put your results in a bigger perspective, I'm going to think of an effort that Nora Eusti coordinates and the commitment to everything, et cetera. This is a problem. The impact evaluation gives us negative, which is horrible. You spend a lot of money. And if I do the distributed analysis, probably this goes to middle class, for upper middle class kids. And they say that if you go to Shail, the channel gets shot. According to more general method on how you allocate public money and public effort. Am I too far from reality? No, no, no. You are completely the reality with your analysis. There is a big debate in Brazil about this program. And there are a lot of evidence also that many students did not study at all. Because there was no rule to the minimum number of courses you need to do. There was no requirement of quality. For example, you need to enroll and also get approved in some courses. And as I mentioned, you are basically giving a lot of money to young people to go to Europe. So if you take, for example, one scholarship that is 1,200 euros to go to Portugal, is two times the minimum salary in Portugal. And maybe the student is more than the per capita income in their household. So yeah, there are a lot of discussion on how the program was created. And yeah, and how was the monitoring of the program and the rules, etc. Yeah, one thing that we cannot observe here is that many students can could increase the cultural capital, the networks. They could influence peers, for example, if they had a brother or sister that saw him going abroad. And so I also want to have an international experience when I do my university. So this could be a positive fact, but we cannot observe this here. But yeah, I totally agree with your comments. I think one, following up on this, I think one important thing that is important to clarify is that your IV is measuring the effect on compliers, right? So those are the marginal students that gain access because of the random compatibility. This program might be good for people at the top, right? Yeah. So I think this is really important for you to sell the paper to make a more general contribution. Because it might be the part of... There will be a group of our stakeholders here, the guys that have super-high interest-exempt scores. And they probably are people from rich families and have very good background, for example. These are not the guys that we are capturing here. Yeah. Good things. It's a long story. The first year that was really hard to get into the program. Because you didn't only have to apply, but you needed to have an research program with a professor at your home university. We revealed a few more students going and students that were committed to study abroad. Yes. On the cost of our training, I know that you knew the universities. So you had these universities that you wouldn't really choose from. So I don't know if this changes your results in any way, but it's good to control. And the third thing is, can you maybe do a regression controlling for your graduation and see the results hold? The university didn't provide the year for graduation. I'll follow up on my previous program. I'm trying to compare the whole set of social issues with the country and some sort of decisions of the margin. And that's what you were writing for countries like Argentina, Brazil, and the United States. These countries don't change. And this case would be exactly the opposite of the previous one. Because in the previous positive effect, the amount of money spent in the policies is probably small because this is a regulation. I don't think there is a margin for this, et cetera. And then in the third or the fourth generation. So just my mind will come and do this. Every paper should finish with two lines consisting of additional aspects usually not considered in the previous paper. I'll follow up on this intervention. It seems that was right. Yeah. Thank you. Yeah. When you have such a controversial policy as a very large affirmative action policy as this one, it's really important to put in a larger picture and have very, I think, convincing rigorous evidence on where it works and where it does not. There are some aspects of this policy that I have worked on that too. Some aspects in the design that can be improved. And we also talk about that. But the overall picture is that this policy has been very successful in decreasing inequality in access to universities. Yeah. Chico. No. The problem is that Prova Brasil for high school concluders was not sensitive until 2017. So, yeah, we have a small sample problem there. We try to use a name. But then there is, it's not ideal either because there is a large difference in take up. So then composition gets really mixed up. So every time we try to do something with the college, with grades, we don't get a clean effect. That is not mixed up with composition, things like that. But, yeah, we are trying to get a sample for which we can do this. But, yeah, we haven't been super successful so far. My other paper. Yeah, I have another paper that shows that people move from private to public schools in response to this policy. So in this time period, that's why also we stop in 2015. Because for this time, yeah, they need three, they have to be in high school for three years. So, yeah, in my sample, there is, yeah, no endogenous starting to high school. But, yeah. I'm not sure it is relevant because there are some sites but maybe also the creation of new companies. So, for example, we can have a university in one state and a university with more countries in different cities. So I don't know how this would affect the college. Yeah, so this expansion of college seats is, I mean, the idea is that the policies are not correlated, right? They are orthogonal to each other. We can't control for this. So we can observe an opening of a campus or the number of seats. We have control for this in one of the specific issues but it doesn't change anything. The idea is that, yeah. And do you know in which states or regions the effect is stronger? For example, if you drop out the data, like you have a high impact, they somehow know that the guys have this communication or clear their name as Mao Zedong or states of North-East that the guys probably have more difficult to know that they have this kind of change in the situation. I know that the larger expansion is actually in the Northeast. So like the South already had quotas, but like the Northeast, I don't know if the effect comes from, but like the large, yeah. The national law affected more universities of the Northeast, but yeah, this is a good point, yeah. Let me see if I have time. Yeah, so I will take the questions and then I answer, yeah. Can't you just compare places where you have good private colleges and then you have an outside option that is good and places that you don't have a good private college? So if the percentage of people not taking the name in those four countries is different, it's indicative that people are just going to private colleges, right? And are you just seeing if they take the name in the year they graduate? Yeah. Can't you just see if they take the name? Because the person that came to my mind, it is the name, like if they can get here or something. So maybe just say, no, maybe this back, yeah, maybe that back when he has it described to me. I'm going to take the questions and then come back, yeah. I don't know if that's correct, but around the same time, it's really popular to stand in the same way that it increases supply of private colleges. So I don't know if the displays need to be private, what would happen because of a crowd out of, you know, actually not a crowd out, but get more supply of private universities. I'll see you in a little bit of time. Yeah. Yeah. 2012. It should be, let's say, control your... Nice specification. Yeah. There was any other question here? Yes. And you mentioned that you were estimating, realizing how for all of the trends assumption, I'm not entirely sure what it's appropriate for your attitude. To whom are we expecting these for all of the trends? Could you clarify that for me? Yeah. Okay. So first, regarding the displacement of private colleges, and I think this is a good idea, Le, to do this heterogeneity, whether they're a good substitute in the private system. So the problem is that to do that, we need the data, the restricted data from SADAPI, so we are still entering with the project there because then we really need to connect individuals and follow them later on. So, yeah, we are doing this, I don't know if it takes time, but I think, I mean, if we want to take this project to the next level, we are going to, yeah, we are going to do that. We are doing that. It's just that, yeah, it gets a lot complicated when you leave the public data and go to the restricted data. Regarding Renata's question about the increasing supply of private universities, I think the answer is similar to my answer to Rodrigo. Rodrigo asks about the supply of public universities because there was a reunion, which was the big program of expansion. And my answer to that is that, I mean, I think this matters for human capital accumulation, but for this to bias my estimates, they would need to be correlated to the exact pattern of adoption of affirmative action. I mean, I think I have to show you, I don't think this would be a problem. I haven't controlled yet for, like, supply of private seats. I think I should do that, yeah. I controlled? Yeah, yeah. Yeah, there were, yeah. We, what we do is we control for, there are a lot of policies for public colleges, like Fies, Reuni, all those things that we try to account for, but yeah, there's also the private universities that, yeah, high education is a fascinating shooting in Brazil because there are, like, so many reforms as you saw. You were right. This is important. And my last answer to control versus treatment here. So, like, treated institutions are the one that expand the most, right? The ones that had the lower level of quotas before the policy, and then they were more affected by the national policy. And then in this case, our treated institutions are not institutions, but micro regions, right? Municipalities, re-aggregating institutions, the treatment is at the institutional level. They treated are the ones that were not affected by the national reform, basically because they already had affirmative action policies before, okay? We also have an alternative treatment of state universities that had no quotas before and so on, but this is our, like, we chose this as our main, yeah. So, yeah, I conclude here. Thank you.