 Hello, everyone. I will ask just an additional minute to explain why I'm presenting now. So I was not supposed to present this paper in this conference. So what happened is that a lot of people couldn't come for the conference for different reasons. So Alei that was supposed to present now, he kindly agreed to move his presentation to Friday for another session. And they jumped in during the breakfast today to present. So I apologize because this is a presentation for 45 minutes. And they just jumped some slides because I couldn't prepare a presentation because we just decided during the breakfast, okay? And this happened when you were working the organization that is part of the conference. So yeah, so this paper, the title is The Impact of the Study Abroad. Evidence from a massive government scholarship program in Brazil is a paper joined with Otavio and Dandre, both from the Jatuljaga Foundation. And I'm Rodrigo from Unwider. So basically, it's a really easy paper to motivate for European audience and present two times already for the European audience. And I basically say, you know that mobility is important because probably you were a Erasmus student. So most of the Europeans in Europe, studying European universities, at some point they were benefited from Erasmus program. And mobility has a lot of benefits for students. It has the benefit of learning different cultures, learning languages, increasing the cultural perspective, learning different kind of label markets, et cetera. And this can be even more important for developing countries considering that those countries have lower level of investments in education and science. And even though many developing countries such as Mexico, Colombia, Chile, Saudi Arabia, China have some kind of international mobility programs, there are no evidence of the impact of this program. In fact, even the evidence for developing countries, for European countries for example, is very scarce and they will talk about that. So just a summary of the paper, so we basically estimate of the impact of this program that's called Science Without Borders in Brazil or Sciences in Frontiers in Portuguese, Enrollment of Masters in PhD programs in Brazil, Formal Employment, Wage and Entrepreneurship. The contribution of the paper first, this is the first paper to estimate causal effects of study abroad programs in the developing country. We also add to a few pool of papers estimating causal effects of those programs. And I will try to convince you that we are the first paper to estimate the effect on those outcomes even for developing countries. So there is no evidence, even for Erasmus. We build a novel data set by merging 17 public and non-public administrative data sets and they will explain how we did that. And we are proposing a UIB strategy here. So, surprisingly, we find that the program had a negative effect on probability of enrollment on postgraduation and on having a formal job and no impact on wage and entrepreneurship. And we try to explain why this happened and we can show some evidence of delayed graduation and the lower probability of brain drain. So, the literature, as I said, is very scarce, the evidence of study abroad programs. The exception is this paper published in age 8, 2011, and they study basically the impact of Erasmus on international mobility. So, the question is basically if a student from Finland go to Germany, they stay in Germany or not. And then after this paper, three papers did the same identification strategy, look into different countries, also in Europe. And they basically use IV that is quite simple, that is if the student is in a university that was exposed to Erasmus in the past. So, for example, if I'm studying a university that some students before me went to receive a scholarship for Erasmus before. But these four papers that try to provide some causal estimates, they use movers against nonmovers and survey data. So, they can only observe from the university students those that were Erasmus students and those that were not. In our case, we will look to our students that applied to a program and compare who win and who didn't win the scholarship. So, part of the selection into the program is already corrected because of four data sets. So, the program was created in 2011. The goal was to send students for six to twelve months for some university abroad, mainly US, Europe and Australia, focus on undergraduate so in five years the program had 73,000 scholarships. To put this in perspective, between 87 and 2000, the same institutions that organized this program, they offered 13,000 scholarships between undergrad, PTD and postdoc. So, the means of science education was responsible for the selected the scholarships, recipients, priority areas and the priority areas were basically engineering, math, statistics, so technology-related areas. More about the program. The program was huge. So, basically every student received a monthly stipend. The average monthly stipend was 1,200 euros, sorry, 1,200 dollars. So, it's quite a lot of money for 18, 19 year old students to go to Portugal, for example, or Hungary, for example. Plus, airfare, housing allowance, health insurance, installation AD and AD for educational material. So, it was a lot of money spent in this program. The cost of the program was 2.72 billion dollars. As a comparison, Erasmus, for all European countries, spent 14 billion dollars between 14 and 20. So, you see how expensive this program was. More about how the program was expensive. The program was five times the average expenditure necessary to maintain a student in a public university during one year. Or I think this is the best comparison on how the program was expensive. It's the same cost of a school meal program that exists in Brazil and attends 39 million children. This is some graphs from a research in Brazil called Fernanda de Negre, basically showing that the increase of spending in science and technology in Brazil between 2011 and 2015, of course, this is not everything science without borders. But when we look to the data of scholarships for undergrad in Brazil, we can see this. And this is science without borders. Okay? So, how we did that? How many minutes do I have? Okay. So, how we did that? First, we built this data set. This data is not public. So, we need to get this somehow. First, we request to the Minister of Education, to these two institutions, the information of all candidates that applied and the candidates who were approved for each call of the program. And I will explain what this call. And they provide as part of the Brazilian Social Security Number. So, the Brazilian Social Security Number has 11 numbers. They provide the six intermediate digits plus the complete names. So, after we had this main data set, we needed to get the data from the universities. So, Brazil has 69 public federal universities. And the Minister of Education is very difficult to get the data for those who are much more than me. That's how difficult it is. So, in Brazil, that is this information access law. I think this is the translation. So, you can send formal request to the Brazilian government and request some information. Whatever information you want and each institution can decide if they give you the information or not. So, from the 69 universities that we send the formal request, we send one request to each university, 13 universities reply and provide information. So, we cover 20% of the sample or 20% of the universities and 20% of the applicants we managed to get basically socio-economic information and other academic information that we need to the, to the, to our identification strategy and mainly the entrance exam score. So, to enter in a Brazilian university, you need to do what we call vestibula. It's an entry exam. It's a national exam. So, and this is really important for our identification strategy and the universities that provided this information. Okay. Then, we have the formal labor market in Brazil that we can follow students in the labor market. And these two data sets are public. You can basically go online and match. But you can match, merge using the complete names and these six numbers from the social security number. Plus, for one specific university, we have detailed data. So, we use these to explore the mechanism in our study. So, this is, there's a geographical representation. So, we cover all macro regions in Brazil, almost all states. So, we have a very good dispersion of universities. This is to show, basically, the candidates. So, we have different majors and the sample for majors vary a lot from different universities. I jump that because of time. So, empirical strategy. So, we basically estimate an IV and we control for the entrance exam score. This entrance exam score is a measure of previous ability. So, it has a bit of a number seven ability, has a bit of preparation, has a bit of socio-economic status, family background, etc. We have gender. We have done to, because some students, they have more than one major, for example. I did economics in 2011, then I did, I don't know, math in 2018. So, we just put this. Then we have a admission year fixed effect, major fixed effect, university fixed effect, and the call fixed effect. Standard errors are closely at the call level. And I will explain what this call is because this is important to explain how we will create an IV for this, a proven variable. Okay. So, the program had many calls. How these calls were organized, and this is very important for our IV here. So, the government come here 2011 and decide, I'm launching a call now. The call was for, I don't know, U.S., and students could apply. But students didn't know how many slots were available. Students didn't know the universities they could go if they were proven. And students didn't know if there would be more calls. More calls in the program and more calls for the same country. So, in this call here, the second call for the program, no one knew if there would be more calls. Only the program implementers. There was no schedule. And we can talk about how the program was created later and was not following the best practices of monitoring and evaluation. So, basically, the students couldn't predict if there would be more calls for the country they want to go. And if the call was competitive or not, because they didn't know how many people would apply, and they didn't know how many slots were available for that call. Okay. So, how the program did the selection. First, the program launched a call. Students apply in their university. And then each university selects the students to send to the Ministry of Education and the Ministry of Science. And the university they apply was basically the decision was based on their GPA at the moment of the course. Okay. And but the universities didn't have a rule. So, they could send everyone that applied. Or they could send 50% or 10%. Each university could decide how many students they would send to the Ministry of Education and the Ministry of Science and Technology. Then, when they received all the candidates, they selected students based on the entrance exam score. And these students could not manipulate, because they did this exam score two years before, one year before, three years before. Okay. But I know that everyone's thinking that we are estimating why you don't estimate a regression that continues to design. Because you know that they are selected through the entrance exam score. Because, for some reason, the Ministry of Education did provide us the score of the last student in each call. And because we don't have data for 69 universities, we only have for 13. Okay. If we had for everyone, we could do that. So, what we do is a kind of leave one out. But instead of leaving one out, we leave 13 out. 13 universities. And the hypothesis here is pretty simple. If one call is more competitive for all universities, when I exclude the 13 universities in my sample, this call will be also more competitive for any university in any of these 13 universities. Okay. So, what I mean is, Rodrigo is in my sample. And Alley is not in my sample. Because Alley is in this 57 universities that they don't have information. But they have information about all students that apply and all students that are approved. So, what we did was basically, if the call Alley applied is more competitive for Alley, and I also applied for this call, this call will be also more competitive for me. Okay. And then we show that there is variability in the call approval distribution, blah, blah, blah. We show that the first stage works. We show different specifications in the appendix. And then we go to the results. So, basically we show that the program has a negative probability. The impact of the program is the students less likely to enroll in the post-graduation program in Brazil. We also find that students are less likely to be found in the label, in the formal label market. When we look to other label markets' outcomes, we basically show a negative effect on job tenure. And this is really related to the mechanism that I will show in some minutes. But we don't find effects on entrepreneurship and these are the main results. So, the program had a negative effect on enrolling in a post-graduation program in Brazil and having a formal contract in Brazil. And those were the main goals of the program. Increasing the enrollment of students in Brazilian post-graduation programs and the increasing workers with higher skills in Brazilian firms. Okay. So, as I mentioned, we have only for one university detailed records for students. We have all the students' history for one university. What we do, we try to understand what is happening. So, when we look to graduation, we observe that those students, they do graduate more. But they do take more time to graduate. So, there is a delayed graduation here. And there is enough evidence that taking more time to enter in the label market in developing countries can have negative results in label market outcomes in the future. Okay. Besides that, so I just, this is more evidence to show that approved students take more time to graduate than not approved students. And besides that, it seems that there was some kind of luck because of economic, macroeconomic issues. So, when we just plot the, we plotted this graph, but in bars, with the GDP variation in Brazil and by state, from that is the state this university is based in, we observe that students are proven they graduated during a recession. And that is also enough evidence, enough literature that shows that graduation during a recession can have a negative effect on the label market. Okay. Finally, I know that everyone is thinking, okay, what if students stayed abroad? Okay. So, we managed to find 64 percent of the approved candidates in any of the data. So, the question is, what are the other 36 percent? If we look to this university specifically, it's called Federal University of Bahia data, we cannot observe here that 20 percent of the students did not graduate until December of 2021. So, for this university specifically, 20 percent of the students did not graduate, so they are still studying. Okay. And why is it reasonable to expect that they're in Brazil? Because you needed to come back to Brazil. So, if you, if you went abroad for one year, you necessarily need to come back to Brazil and stay in Brazil for one year. This is one of the the rules of the program. If you didn't stay in Brazil for one year, you need to give the, to give the money back to the government. And that also makes sense that students will come back and finish the undergrad first and then go abroad because they would basically lost all the two, three years of undergrad they already had before they, they, they leave the country. So, 36 percent, right? So, they can be finished in the undergrad. I know that UFBA is 20 percent. They can be unemployed and looking for something or they can be unemployed by choice and of course they can be abroad. Okay. So, however, there are mobility constraints to Brazilians. Why? Brazilians do not have work permit in Europe, US or Australia. Also, there is a cost to go abroad. It's expensive to apply for a visa, to get a job, to rent a house, etc. Also, Brazilians need to pay high fees for post-graduation programs in Europe and there is no students law in Brazil. Okay. But you can still say, okay Rodrigo, but there is, there is still can be some students abroad that and this could be driving your results, etc. What we are doing now is basically we are doing a scrapping on LinkedIn and try to find these students on LinkedIn. These 36 percent students because basically there is no data to find those students if they went abroad or not. Okay. Thank you.