 Thank you very much for this invitation. And this is a joint paper with Jorge Rodíguez from Monizia, Chicago, and Sergio Ruzva from Maryland University. So let's see, here we go. As you might know, Chile is one of the most unequal country in the world. So here you have the countries in the OECD group. And as you can see, Chile ranked as the top of the countries with very high income inequality. And here you have a similar information using per capita income by some time using Cacen, which is the Chilean household data set in 1890 and 1911 in real pesos from 2011. And this picture is quite interesting, in my opinion, because if you exclude, for example, the top 5% of the distribution in Chile, Chile is quite an equal country. And the genico-efficient is about Germany. However, if you include this top fraction of the income distribution, Chile became one of the more unequal countries in the world. So basically, the action is here. If you want to understand the inequality and the pattern of inequality in Chile, you have to pay attention on what is happening in the top of the income distribution. So here is coming. This picture is coming from this project at Kinsom, Piketty and Saez, and showed the participation of the top 1% of the income distribution in 2010. And you see here, Chile is about 25%. The top 1% concentrate about 25% of the total income in Chile, which is extremely unequal. We already learned that Ecuador exceeded 20% here, and Denmark, Sweden is about 7% over here. So in any metrics, Chile is very unequal country. So the key questions for these papers are, what are the underlying causes of this? What is the role of the schooling system and how educational policies impact in any way in the labor market later on? So this is the basic question from this presentation. And in one slide, I will try to explain how the Chilean educational system works, because it's gonna be important to understand the next issues in this presentation. So in 1981, the military government in Chile established a voucher scheme. And as a consequence of that, nowadays, we have three types of schools. We have public school, voucher school, and private pay school. And voucher school concentrate about 54% of the total enrollment in Chile today. So in addition to that, the voucher school are allowed to charge a co-payment for families. So you apply for this kind of school, the school in addition to the subsidy provided by the state, the school can charge an additional fee to the family. Also, this type of school are allowed to select students. So usually this kind of school, what they do is to select the smartest students and also the richest one, in order to obtain the better student to be educated. Finally, these schools can run for profit or not for profit. So there is a huge discussion on this issue today in Chile, that is educational reform and the discussion. But anyway, this is the situation today. There is a large evidence on the school choice and educational achievement using data mainly on education. And according to this data, public school and voucher school are quite similar in terms of educational outcome when you control for family background and income and stuff like that. And there is a clear advantage of the private paid school with respect to the public one. However, this previous evidence is limited by data. Most of the data is coming from cross section data. So it's very difficult to properly identify causal effects. This table is to show you a little bit about the consequence of this educational segmentation. Here you have public school, private voucher school and private fee paying school. And these are the standardized tests in language and in mathematics. And as you can see here, this test score increases according to this type of school. So public school obtained like 240 points in this test in language, private voucher school 256 and private fee paying school 275. And the same part of it is show for mathematics. So in this paper, we're going to use a new data set for Chile with a panel data. And basically we are trying to argue in the paper that because we can control for pre-labor market abilities and individual social characteristics during the high school, we can better explain the contribution of each type of school in terms of, not now in terms of educational outcome, but in terms of impact or outcomes in the level market. So we argue that we have here a very identification strategy. So the main results, you want to leave now. This is the main result. We find a clear link between individual high school type, public voucher and private and the level market outcomes. Particularly private fee paying school have a higher returns in the level market later on. We also are going to mention a little bit about two educational policies which are very important in Chile. The one is the extension of the school day and the other one is a teacher incentive. Both of them try to improve and increase the quality of education in Chile. However, we found no effect in terms of level market outcomes, okay? So we spend a lot of money with no real impact in terms of outcomes in the level market. So papers investigating income inequality, there is a lot, most of them cross-sectional data, more recently some cohort studies, but not much using these kind of panel data. So this is the first paper trying to link data on individual schooling and level market outcomes. We argued that this allowed us to study the origin of inequality for a recent cohort and this paper follow a previous literature, mainly done by Jane Heckman from University of Chicago. So you can see the more, the empirical strategy and the algebra in the paper with more details, but at the end of the day we come out with a model like this reduced form model in which we are going to explain a level market outcome which are earnings, okay, in period T-bar, okay? Period T-bar is going to be 2011. Thus, we're going to observe earnings in 2011, this T-bar. And this earning is going to be a function of a vector of exogenous characteristic, school characteristic, family background, individual's ability and public policy. All these coverage are going to be measured at a particular period before 2011 and in particular is going to be in 2001. So we have a panel data, same individual 2001 when they were 15 years old. We have all the family characteristics. We have an ability test measured by this standardized test in Mathematics and Spanish. And then we're going to observe the same guy, same individuals in 2011. You may believe that education in terms of level and the school choice is not exogenous, actually it's endogenous. That's the problem with the previous literature because they are using only cross-section. You may think, for instance, that the weather families or a parent with more education or maybe the school choice is correlated with the ability of the students. So all these factors can be controlled in this analysis because we are controlling by family background in the past and also for individual abilities. So we can reduce substantially the potential selection bias by using panel data. I mean, exploiting these cross-section variation and also the time series variation. So let me talk a little bit about the data. The data, in the data we observe the test score at the age of 15 and this information come from the measurement system for education quality sim set which is the standardized national test, okay? So this test, we have this data set for 10 graders, people at the age of 15, okay? We define our exogenous characteristic vector QI including age, age squared, gender, previous attendance to early education, FI include mother and father education, family income and number of vocals at home. And these abilities include the language and math test score and we also have a variable indicating if the student to get a previous course. We observe student earnings 10 years from the time they took sim set, I mean in 2011, okay? And we link the educational data set with the unemployment insurance database using the national identification number for each individual. So we were able to match the data, okay? And in this way, we build this patterned data. So, of course we have problems. I mean, that's an initial case for empirical work. We have sim set database for this sample, which is, that's the sample for students. However, our analysis is based only in 78,000 individuals. So just to clarify the situation, we drop the student from the database with missing values in some of the coverage from sim set database. Then we consider only student affiliated to unemployment insurance system which are formal workers, okay? So this study is gonna be relevant only for formal workers. We don't have informal workers at the age of 25 or 26 years old. And we leave also information with not zero total 2011 earnings in our final sample. So I know you might be worried because of all this selection process. However, as you can see here, this is the original data with the whole sample and this is the sample that we are using actually here after linking the data. And as we expect, the earnings are much higher for this group than for this group because we're selecting people who are actually working in the labor market. So this is a problem. However, in terms of the other coverage, age, the mean for this group is 26 years old and for this sample is 26 years old, okay? And the test score in language is 251, 251. In math, 246, 247. Percentage attending public school, 48%, 48.6 and so on. So pretty much the sample are no bias because of this selection process. So just to motivate the discussion, here you have the model education and the relationship with the simset, the score and future earnings. And as you can expect, model education is positively correlated with higher results in math, language and earnings. So here you have the score when your mother has primary education to 30 points and almost 300 points when your mother has university education and the same pattern is for language and the same pattern is observed for monthly earnings. Here you have the non-parametric relationship between the mathematic test score. So that's the score of the test. A good score is about here, like 350. That's a pretty good score, okay? And this is the earnings from the same individual. So here you have the panel later, okay? So that's the score in mathematics. That's the earnings from the simset of individuals. This is when this individual has a 15-year soul and this is when they have a 26-year soul, okay? So we can learn three things from this figure. First of all, there is a positive relationship between mathematics, score and earnings. So math is quite important. That's a good thing, all right? Second, there is no much difference between public school and voucher school. You see? I mean, these two lines are very close. There is no much difference in terms of future earnings. Finally, there is a significant difference between these two types of school and private fee-paying school, especially for the highest score in mathematics. You see? I mean, the action is not here. Most of the action is about here, okay? And this is quite interesting because if you want to enroll to your kid in this kind of school in order to gain this gap, these types of school are quite expensive. And this is going to be a key issue for the conclusion of the debate. So if you run the regression, the earnings regression you can see here, I mean, here you are controlling, let's concentrate here in a specification number seven, we're controlling for all the external characteristics we mentioned before, family background, previous performance, policy on levels, policy with interaction, et cetera. This is the sample size. And we can see here that if you attend to a voucher school, you obtain almost three extra points, 3% in terms of earnings with respect to attending a public school. If you attend a private fee school, you'll obtain 15% more, so a high return to attending a private fee school. And also, this is interesting, I mean, the gain for a higher scoring language is 1.4% and for math is 14%, okay? So that is a much gain for an extra point in mathematics rather than in Spanish. So when you control by the average at the school level, try to capture something like a peer effect or so on, you can see that the voucher school, the result are quite similar for the voucher school is the gains for attending voucher school is about 4% here. And private fee paying school is 7% rather than 15, but you have to add the effect of the school level, I mean, for the peer effect. But pretty much the conclusion on the difference between public and voucher school remains. I mean, there is a small gain in the voucher school with respect to the public school. So then we examine two educational policies. As I mentioned before, first of all, the full school day program. So there was an increase in the school day program for the kids. They usually attend from eight o'clock to 1pm. After the reform, they attend from 8pm to 4pm, okay? So there was an increase in the number of hours of lecture at school. That's the heck of full school day program. And the second was the national system for the school performance assessment, which is a teacher incentive. If the school performs better, then teachers are paid a bonus, okay? So let me skip this. So when you analyze the effect of the extension of the school day, you see no effect. The parameter is not significant in term of future earnings. And also when you interact, the extension of the school day with public school, with voucher school, that is no gain. The only gain, again, is obtaining but student attending private paid school. When you analyze the same for the teaching incentive scheme, that is no significant effect. Only you observe a positive and significant effect when the school earns these bonus three times in a row, okay? We represent about 3% of the schools. So again, this policy does not create a big change in terms of outcomes. So when you interact, the incentive scheme with type of school, again, the only benefits are kids attending voucher school when the school was benefit by this scheme three times in a row. So let me finish before the, you know, the rain, exactly. The main conclusion is when you control for extension characteristics, ability and family background, we document that different type of school produce different future labor outcome on a student. Most of the action is given by private high school with more than 300 points. We show a higher return to educational expenses and that is something like an intergenerational transmission of inequality. Elites begets elites, okay? That's pretty much the conclusion. And in addition to that, educational policies are not working. They're not working in terms of closing the gap between public education and private education. Thank you very much.