 Hello, hi everyone. Thank you for being here. I'm very happy to present this joint work with Andres Moya and Fabio Sanchez, who are two professors at the University of Los Andes, just like me. And Fabio is here. So there are some very political contextual questions. He will be the best person to address all these aspects. So what we do in this paper is to study the Serpilopaga, which is a scholarship that many people, Colombians know, at least, that was paid during four years for the University of their choice for 10,000 students. These students were selected based on two criteria. They needed to be below a certain level of income and they needed to have a really good grade. This was a both merit-based and need-based scholarship. And we will see the question is that this increased inequality of opportunity because now if you're poor and you do well, you're able to go to university. So we see this as a real change in opportunity and I won't show you why. Does that translate into an increase in human capital and we're especially interested in not the exposed effect of the scholarship, meaning those who receive the scholarship tends to be able to go to university, have higher incomes. All this has been demonstrated pretty well, but we know much less about the fact that this opportunity exists when you're a college student, how much does that motivate you to study more and get better grades. So there is a first motivation which is about what are the additional benefits of a scholarship and as I was saying, we typically think of benefits under beneficiaries, but when you want to think, is this kind of scholarship worth it or not, you also need to think about this motivational effect. And as we will see, there's a whole lot of students that improve their grades just because they know that now if they get good grades, they can go to university. And so that's the first motivation for it and one way of showing this is, look at this is assuming that you have a very good grade, you're in the top 10% of the distribution based on socioeconomic stratum, which in Columbia sort of is a good proxy for how wealthy you are. Six is the wealthiest and one is the least wealthy. Then how likely are you to enter an accredited post-secondary enrollment? Meaning an accredited university, meaning a high quality university in Columbia. You're less than 20% likely, even though you get a super grade, you're less than 20% likely to go to university if you're in a socioeconomic stratum one or two, but you're super likely to go, well almost 60% if you're in the very rich. Once the scholarship appeared, this is how it looks like. So it really aligned the, now if you have a really good grade, you're as likely to go to university if you're poor as if you're rich. So this is one of the most drastic change in opportunities that we have ever seen. And hence the question that we ask is, this is already a result from another work from Fabio and Juliana Londoño and Catherine Rodriguez. So our question is how this change that we see, this change in opportunity actually changes the motivation and that's the results of the students. And the best sort of, the second motivation does is really why social mobility is so important. Social mobility triggers human capital. It's beyond a debate on just education and is scholarship worth or not is the effect of social mobility on human capital accumulation. So on the, I'll skip this slide for the time constraint. So let's summarize what Serpilopaga is. It has 10,000 new students per year and that costs about 2.7, that means 2.7% of all students. It funds entire undergraduate students, typically four years of education. And the downside of it is that it's very costly. The first year it was 4.6% of the budget of the Ministry of Education. And since there was a new cohort every year that meant 20% by 2019. It has two criteria. One is the need-based eligibility, which means you need to have a CSBEN score, which is a social economic index, below a certain level. And basically we estimate that this is, mean you need to be in the 55th percent poorest of the country. You don't need to be super poor in that sense, but in relative terms for Colombia, but it's this year of the population. And merit-based. And basically you have a certain score and the score is set so that we have about 10,000 students per year that could enter it. And the timing here will be very important. 2014 was the first year, but this was decided after the exams. So this would not affect the motivation of the students. But then everyone, it has made a lot of noise in Colombia. Everyone knew that there is this huge education program that promotes the best performing poor students. And hence in 2015, it's fully credible and there is a full effect on the motivation of the students. And that's what we're interested in. Okay, now what is our prior and one sentence is that it's the top of the distribution. In Saber 11 is the end of high school test that we have in Colombia. It is the top of the distribution that should be affected. If you're below median students, you know you will not get at that level in one year. You cannot make that huge jump. So you won't necessarily change your level of effort. But if you're a top, if you're already a good student, then you should be the ones that are most motivated. So in a descriptive table, this is a sort of difference in differences that is both sort of, it's useful to give some stats, but it's also useful to get a sort of a difference of indifference of a look at the result. If we look at the average in the difference in difference, basically this is a rank. Think of a rank of the population. If you're zero, you're the lowest. If you're a hundred, you're the best student in your cohort. So if you're in the eligible ones are the poor ones, there's 55% poorest of the population. The average rank is 45 versus the average rank is about 55.2 for the non-eligible, meaning the wealthiest one. So it's a 10.4 difference. This difference, we will call it the socioeconomic achievement gap, 10.4. And this we can look at it percentile per percentile. This sort of analysis by percentile will be central in our analysis, given the hypothesis that the top percentile is the one that should react the most. So this is the initial gap. And we look at how it changes from 2013 and 2004 to the year 2015 with the motivational effect. And there is a drop of about, here it's a small drop in the average gap and the biggest drop when we look at the differences in ranks from 9.7 difference to 8.8. And this drop, we basically divided by the initial value to get this can be interpreted as what is the gap as a share of the achievement gap of the initial achievement gap. So there is a 4.6 percentage reduction in the achievement gap in the entire population on average. And there is a, this is concentrated in the top percentiles. That's where we see the reduction, especially at the 90th percentile, there is a 9% reduction in the initial gap. Meaning the students, they didn't receive, and this is still somewhat descriptive, that that's not our best estimation, but the message is they didn't receive more books, they didn't get more buildings. The only thing that they got is the hope that if they have good grades, they can go to university. And just this reduced by 9% the initial gap, the huge gap that there is between a wealthier and poorer students in Columbia. So the initial strategy that we had, in fact in the previous version of the paper was just to do a regression discontinuity. We make use of the fact that there is this ceaseband threshold where the students that are slightly poorer are eligible to the program, those that are slightly richer are not eligible to the program. So this is perfect for doing a regression discontinuity. We had the detailed data to be able to do this. And so in this you have this effect of being eligible, controlling for the ceaseband score and interacted with other variables, et cetera. The reason why we cannot do this, and we got into a problem program when we sent it for publication, was that there are a number of other programs that use the same, these are the threshold, the threshold is different if you're in the 14 cities or urban or rural, but we normalize to take this into account. The reason why we could not do a simple regression discontinuity is that many programs share the same threshold. Some of this are early childhood programs, a housing program, even credit for students, et cetera. So the main argument, counter argument that we had is if you want your regression discontinuity, the logic of a regression discontinuity is I want to compare students just below and just above and below the eligibility threshold. It doesn't work anymore because now we have a difference that could be driven by the other programs that are also triggered by the same eligibility. So what did we do? You do a difference in discontinuities. This is based on the fact that most of these, all these programs were pre-existing. Hence, now what we will do is think of it as we do regression discontinuity in 2015 and we do a regression discontinuity in the years before 2015. And we take the difference between the two regression discontinuities. And that is the change that happens as the threshold. So that could be attributed to the program. And that is the regression. We have this eligible effect, but we're also taking out times 2015. Now, these are the results from the regression discontinuity and basically now we see even stronger results than when we use the difference in differences. And what we find is there is an effect that is really starting at the 75th percentile and above and the 90th percentile. At the 90th percentile, meaning the pretty good students in the distribution, we find the 1.57 change in the ranking and that represents 16% of the initial gap that there was between the rich and the poor. Now saying without giving them anything else but the opportunity to do better if they enter to get to university and they have better results, they increase their results by 16%. So that's probably the key result of the paper, but we have other. So if we look at a percentile, now we do this quantile RD, different RD at every level, we find that there is not much happening in the bottom of the distribution as we hypothesized at the beginning. But if we look at the top of the distribution, there is a significant and positive effect that is happening for these good students at the top of the distribution. The typical discontinuity graph tells us that, again, just a quick look tells us that there is not much happening in the bottom of the distribution, but by the 75th percentile, we do find something. And when we look at the 90th percentile, then we see even more substantial and clear differences between the two groups. Now we want to say, so I won't have time to re-summarize what I said, but now we have additional results. One is looking at now sub-air nine results. So now instead of being an 11th grade students, we will look at ninth grade students. The data is less detailed. We cannot do regression discontinuity there. We have to do just a difference in difference by school based on how many eligible students. The result to summarize this result here, we find that once the CERPILO-PAGA program starts, the schools that have a higher proportion of eligible students and start having better results compared to the schools that have a, in change compared to the schools that have less eligible students. And that means there is an improvement that happens among schools that have higher eligible students. And that is also likely to be attributed to the demotivation effect. Of course, it's not as well identified as the regression discontinuity. The other, the last question that we try to answer with the data is, how much is this now affecting the enrollment rate of the students? So we know something already is that the top 5% of the students here, so now we have a curve that's comparing eligible to non-eligible students and 2014 to 2015. What the data tells us is that if you look at the eligible students since 2014, they have a clear jump. And this is a bit like the first graph that I was showing you. There is a very high level of admission for these students. And this is because they're receiving the money. Okay, that's what we would expect in 2014. They're receiving the money they enter. But this chunk of students that improved their grades substantially between those that were just between the 70th and 90th percentile among the eligible ones, they improved their grades. We know this. And now with this graph, we know that they also substantially improved their likelihood to enter university. Hence the fact that they improved their grades even though these are sort of the disappointed ones, they worked hard, but at the end they didn't get the scholarship. And despite the fact that they didn't get the scholarship, the fact that they worked harder and they got a better grade, which is also a criteria for entry at university, they were more likely to enter university by some around five percentage points for most of this group. And so that is another key result. And that tells us that there is a substantial increase in human capital that results just from the motivation. This is not the money effect. This is the effect of having worked harder because of the possibility of receiving the money. So imagine how a society that has better equality of opportunity could actually change this level of motivation. Just, we all work not only for publishing papers, but trying to influence some of the policy makers. So I thought there was a useful quote from President Santos when he was president, where at some point, when he was defending Serpilopaga, he said, the program affected the quality of education through a number of things. And then the motivation of the youth put more effort so that they can access better universities, a question that was reflected in the recent results in the cyber exams. So we were pretty happy to see that it reached all the, so Fabio, we shared the results and the lessons. And when the debate was made, it was around should we continue Serpilopaga or not? And we know that it did not continue as it is, it became generation E and lost a lot of its magnitude and impact with it. But it did have some weight on some of the decision. To conclude, in the bigger pictures, what we see is further that simply the motivation generated by Serpilopaga increased grades in the top of the distribution by 0.09 standard deviations and that represents 16% of the initial social economic gap. The student's engagement is really driven as a key input. And this key input depends on how much opportunities do they have. But this effect is concentrated at the top of the distribution so we should not think that it's a substitute for better quality, better teachers and all the other. It's something that may be a compliment but it's not working for everyone. So obviously it cannot be a substitute. So in this debate, but is it worth it? We just say there is another argument that needs to be taken into account when we think should we have this kind of scholarship or not. Thank you because time's over.