 Hi, it's really a pleasure to be back here. I'd like to thank the opportunity. So similar to the other papers, we have tried to solve the same questions, more or less. Have a very good team of people working this project, especially some young guys, which are starting to research. I think that was a good practice and a good part of the project. So talking about the Brazilian case, I think the spirit of the project reflects very much what Peter said about we have some new data that makes us possible to fill some gaps that we have in the literature. So that's very much the spirits of the project. So let's start with a picture of per capita income in a quality in Brazil. So it was quite high, not much trends, a lot of instability. Then in the beginning of the millennium, or perhaps in the mid-90s, it started to fall, a very consistent fall, like a genie fall of more or less eight percentage points, eight genie points in the spirit. If you look at labor income concentration, a similar trend. So now Brazil inequality started to move. If you see that the series really started in the 60s, where we had a big rise similar to that decline. So that's basically the story. I think to describe, well, what happened in Brazil, one should also look not only to equality or inequality, but also to growth of household income. Because this inequality fall is very common to what happened in many Latin American countries. There are some work by Nora Lustig, by Leonardo Gasparini, that say, wow, that's a general trend in the continent. Some sort of a convergence of inequality. Latin America is very unequal. Inequality went down during this period. If you see the period of inequality fall, starting 2001, it was also coupled by growth. We had a commodity boom. We had a growth period, not as much as China, of course, but for us was quite a good growth. But a kind of growth that it reflected very much on household income. So we had three components. If you decompose the growth of median income or social welfare, you have three parts, more or less equal weights. Inequality fall, one-third, second, GDP growth, and then household income growing above GDP. So these three components. So this is the story until 2015, when we see a start of a big recession there. We see this big fall of income. Inequality stopped falling. So try to see some more recent data. So there is a problem here and also an opportunity that Brazilian data sets are changing. So Carlos will have a lot of work ahead. Try to reconcile these different instruments. And so we have here the growth rate of median income on top of that of equality and then the social welfare. So these were the good years where everything was pointed to the right direction. They're in 2014. More or less when I came here first time, you were in the summit and then start a long, long fall. So basically, this is new data. It's not a consensus. They are just for labor income, this part. So basically, after 14 years of falling genie, mostly followed by growth, we have 11 quarters of inequality rise. So inequality is in a big rise. So Brazil case is interesting either for the falling trends, but also now for the rising trend, I will focus on the falling part where we have more data to talk about. So it's very, very worrying. So even the recovery we are starting now, it's the biggest documented recession in Brazilian history, this. And because normally the recessions are not very long lived and the problem that is followed by inequality increase. So the social welfare effects are much smaller than it would be otherwise if inequality was constant. So basically what we do in this project, we have these different papers that I'm going to show in a second. But what we pursue are two common things in all papers. The first thing, we look at inequality, but we also look at the mean income. I think we have to look at both at the same time. Why? So that's the first characteristic and explain why. The second one, we look at very much at the rates of change. Look at the picture, but also the dynamics, the change. So this project is very much change across time. Why this may be interesting? First, because we're going to see a lot of measurement error issues when you combine. So if you look at the second moment of the distribution, if you also have to look at the first moment, you have more evidence to learn about what's going on in the data. So I think this is a key thing. Second, you want to talk about social welfare. I mean, something, look at the big picture. So you look at the both sides. It also makes it easier to compare different data sets, because as you're going to see in a moment, the problem that is different data sets cover different moments, different periods in time. So it's kind of a messy. So if you have rates of change, it's more comparable across different data sets. So that's what we basically do. We look at the inequality or inequality and the mean, and look at the picture, but also the movement of this series. So this is a project overview. We have the syntax paper plus these six papers. I think it's better not to go into the details now, otherwise I won't have any more time. So I'll try to explain each main result of each paper. And then you have a complete picture. So the first two papers, they use a data set that's not very common used for inequality analysis, which is an administrative record from the formal labor market. It's a matching employer-employee survey. It's a very rich source of data set. We have 33 million observations each year. So it's a huge data set. And when you look at this data set, first when you compare Gini trends, although it's only formal sector, et cetera, they're very consistent what you see with household surveys. The thing what you see is that as we move from the bottom to upper deciles, the growth rate reduces up to the 90% tile. But then if you look at upper percentiles, it kind of bounces back. So it goes from the first to the 9% tile, and then the growth rate is bigger for the top 1% or the dots, 0.1%. So it's your second nice thing about this data set is that it allows to estimate firms' fixed effects. So what are the impacts of individual firms on inequality? And it turns out to be quite big effect. So there divide a population by education groups. So basically you have demographic components, explains a typical regression model, explains something like 20% to 30% of the variance of logs. If you add sectoral occupation on top of that, then you explain something like 35% to 42%. But when you add the firms effect, you have a big jump on the explanatory power. So basically firms' fixed effects explain if you do a tile decomposition, explain like more than two times education. So I think this is a key result there. We do the second paper using the same data set and analysis of gender gap. So basically the same statistics for the gender gap. So these model, these three stages, if you run them, do quite a relatively good job to explain gender gap in Brazil. So I think that's the main conclusion. You can understand what's going on with gender gap with these variables, especially these firms' specific factors. The second evidence on the gender gap. You know, gender gap is falling in Brazil across time. As you can see, older generations, the gap is bigger than younger generations. The curves more or less dominate one another. But there is this life cycle trend until the age of 40. You know, gender gap increases and then somewhat decreases. Something that you find in the literature for German, for the US, a similar pattern, OK? So then the third paper. We also address intergenerational mobility of education. We take advantage of special supplements in the main national household survey, which is nicknamed espionage. So in the beginning of the falling of inequality, in the end of the process, we have these two pictures. And so basically you have the education of the parents when the individual was 15 years old at that age. So basically what you see is that the intergeneration persistence of education from father to son, the coefficient in 1996 was 0.7, which is huge. And now it's 0.47. So there was a reduction in this coefficient, but 0.47 is still very, very big. Only Colombia is on our level now, OK? So it's very big. So Brazil is like in the fourth division. Now we are kind of in the third division. So but what the data show, if you interact father's education and cohort effects, we see that this process has been going on for a while, OK? So it's not like something that only happened there. It's you can track this downward trend cumulatively across generations, OK? The other thing, we take advantage of these data sets. So we are in the part of the earnings inequality. It's to take advantage of these parents' background education in terms of the relationship between individual earnings and education. So the basic other evidence is that the returns of education fell, which is in the opposite direction to most countries. But it has been going on for a while as well, OK? So it's not like a generational thing. It's a cross-time. Brazil in 1980, the mean level of education of the population is like two and a half years of education. Now it's something like eight, nine years, still very low. But it was a jump. And then the returns of education fell during this period. One I think a key evidence that we found here is that if you estimate the returns to education with and without parents' background education, makes the whole difference in terms of not only the levels of the education returns, education premiums, but also the trends. The trend is much clear, a fall of education returns if you take into account parents' background education. Makes the whole difference, OK? So it's like a six times bigger fall. So we also take, besides this omitted variable bias, we also try to take into account the attenuation bias due to measurement error. Take into account who answers the survey. You know, if the person answered himself, then it's more precise. Or if someone else in the household answered that. So we have also evidence of attenuation bias between 14 and 32%. Then the fourth paper, that's the question. Does missing income affect distribution trends? So in Beja, we have a very good statistical office in Brazil. He does a lot of imputation, imputation exercise on all data set except Pinaj, which is the main one. So we had the opportunity to say, OK, let's do that. And the answer to this is no, it doesn't affect. So it's quite a relief. Carlos can relax, OK? So missing education, missing incomes doesn't affect in the end. So we develop an imputation procedure with Pedro Silva, very interesting. Ta-da-da. We're also using puted rents, ta-da-da. But it doesn't affect. So go to the fifth paper. That's, I think, a more substantive, big picture. So what we do there, we do this micro simulation exercise using different surveys, four points in time. And we construct the static picture of not only mean income in a quality and social welfare, but it's changed. So it's like little pieces. You can decompose the trends. And so one mission we had that Fin gave us is to estimate everything using disposable income. So first lesson is that doesn't make, fortunately, a lot of difference on the trends or on the level if you're using disposable income or gross income. It does affect a lot if you use final income or market income, like if you, from 2003 to 2015, disposable income, social welfare, was going 4.8. If I use gross income, 5.1. So relatively close. And the component is like 0.13 or 0.12. Not very different. But of course, if you look at market income or final income after government taking into account also indirect tax, then it makes the whole difference. Looking at disposable income, basically social welfare growth, 72% is the mean, 28% is inequality fall of these trends. So very briefly, cash transfers help inequality, help social welfare, and the movement of direct and indirect tax go on the other way in this spirit. So if you look at the most best well-target program in Brazil, Bolsa Família, it has, each hell you spend in Bolsa Família has 120% bigger impact than the second best well-target program. But Bolsa Família is the concentration curve across time as it expanded, become less and less well-target as it expanded. The net effect on inequality increase. But the concentration has the inequality reducing effect of Bolsa Família itself, the inequality within the program has decreased across time. So trying to get some messages before the last paper, first message, disposable gross income, they're not very different. OK, so role of earnings explains 80% of mean income growth and about 53% of inequality fall. OK. Another point that I think is key in the terms of Brazil that I mentioned before, there is this big gap between GDP and household income. It's like a 2% point a year during this boom year. So since Brazil didn't grow that much, 2% a year is a lot like these two here. It's a lot. And this is a trend that's not happening in Latin America. Brazilian household income growth, if you get the 17 countries in Latin America, is number three. If GDP growth is like number 10. So it's not a Latin American thing. It's a Brazilian, what you say, jabu de caba. It's something that we only have in Brazil, OK? It's like a fruit. But fortunately, we are able to explain this. Because of course, you have a problem. If you compare exactly the same thing in national accounts and household surveys, there is this gap in real terms. But it's because of the deflator mostly. So nominal national accounts and nominal household surveys, they are very similar trends. The problem is that the implicit deflator or the price index, which makes us relief. So it's not the end. And when you decompose these price effects, it's not so much terms of trade. It's a difference between the CPI implicit and the implicit deflator for consumption. So it's then finally, the last point, roll of top incomes in the household surveys in this period, as you move from the top to the bottom, the rates of growth reduces. OK, the bottom to the top, it increases the rates of growth as you move to the top part. So finally, the last paper, we do this Pareto interpolation to combine household survey and income tax data, only publish tables, not the micro data itself. So basically, you substitute the household income for income tax after a point. We do robustness exercise to do that. And here, in terms of rate of change, 2007 to 2015. So basically, what we see there is the following. If I see the Gini with the combined series is like 11% higher, which is a lot, it decreases much less if you use the combined series. But mean income growth is 35% higher if you combine the series. And it grew at a much faster rate. If you believe in this combination exercise, Brazil lived an economic miracle in the end of the 2007 onwards, which I would love to see that. But I don't think it's true, OK? So I think this is a good exercise. When you use this stopping in terms of movements, it's like Brazil, it's like booming. It grew from 2007, 2011, 10% real terms a year, while GDP grew 3%. Which country would you like to live? China, I think, would like to live in the combined series. But maybe it's a fiction, I don't know. So it's not only that social welfare is bigger when you use the combined series, although inequality is much higher and has decreased much less. But there is a Pareto dominance in the picture by construction and in the movement, which could be the other way around. So basically, we are able to explain these differences, taking into account, once again, deflators, OK? And formalization process. So when you come in Brazil, the formalization grew 2% points a year for like seven years in a row. So you are comparing oranges with apples, in a sense. You're not comparing the same guys when you do this combination exercise. The problem, just the last point, is the following. You know, we normally use income tax data because you say, OK, if someone declare it has some income and it's going to pay tax on that, why someone would overestimate income? So it's a good data. It's incentive compatible to use this data. The problem that the main growth we had was on exempt income. And when you look, making a long story short, when you look at the change of the rules of income tax in Brazil, personal income tax, you have basically a movement like, for example, if you see income tax declarations, the Brazilian population become younger when we know it became older. So basically what happened is that people started to declare their parents and their grandparents as their dependents. So there is a huge, there is a confusion there. OK. So I think it's a bit risky to conclude using this combination exercise. I think it's a very good avenue, but we should be, you know, when countries where it's a lot of institutional changes in the rules, in income tax rules, et cetera, you're comparing really different things. So I think my time is over. So this gives you the big picture of the project. Thank you.