 is to give you a little bit of an overview of what happened in Latin America over the last, right before the pandemic in the first two decades of the 21st century. And highlight some stylized facts that I think are important. In a way, this presentation is a little dated. We are going to be listening probably over the next two or three days what the pandemic has done on inequality. But the topic of this wider conference is about reducing inequality and I believe there is something important to be learned from a very big example of a great inequality reduction in Latin America over the first two decades. So this is the inequality reduction I'm going to be talking about, I'm going to be focusing on wages. I'm going to be focusing on labor. And, you know, no matter how you measure it in a quality decline in wages, and here I'm loosely talking about wages. I'm including here also the income of the self-employed. So this is wages and earnings from labor. Although the income of the self-employed sometimes is a mixture of earnings from labor and capital. It's difficult to separate for simplicity because I don't have a better way to deal with this. I'm going to treat those incomes as labor incomes. I'm going to put them together. And this is three different measures of inequality. This is a well-known graph probably. What you see is a very massive decline of inequality that starts around 2002. We end, these are 15 Latin American countries. These are unweighted averages. If you weight, things are a little different. But the main message stays no matter how you measure inequality in wages decline pretty dramatically. What I'm going to be trying to do is to review some of the stylized facts behind this reduction in inequality in Latin America. And I'm going to emphasize two aspects that I don't necessarily think are the most important aspects, but there are aspects in which I've been working on, which are labor supply trends, firm dynamics, firm characteristics. I'm not going to be saying anything about very important elements of the decline of inequality in Latin America over the last 20 years. Before the pandemic, I'm not going to be saying much about minimum wages, the commodity boom or other macro shocks and whatever happened with technology, polarization, et cetera. Not because I don't think these are important forces, actually they are because simply I don't have much to say or much to add to the discussion. This presentation is based of joint work fundamentally with Joanna Silva, who is at the World Bank but definitely work also with Manuel Fernandez Sierra who is sitting here. Marcela Slava, two great economists here at the Uni Andes, Chico Ferreira, Sergio Firco, who are also sitting here and Alvaro Garcia. I'm going to be using data from SEDLAC, much employee-employee data for three countries, Brazil, Costa Rica and Ecuador. And I will say a couple of things about manufacturing survey data towards the end of the presentation for the case of Chile. One of the, I think, key stylized facts that one needs to keep in mind, and I, sorry, these are very, very simple things but I think are important, is that the inequality decline took place because fundamentally wages at the bottom grew really, really fast. Inequality can decline because the wages at the top go down and they get closer to the wages at the bottom. No, this was a period of growth, of wages across the board, but much faster wage growth among those people who are at the bottom of the wage distribution. So it was a very pro-poor growth what took place in Latin America. Again, all of these graphs are 15 Latin American countries, one can do it, these are averages, one can do it, country by country. The results are pretty much the same, the broad picture is the same, there are, of course, some important differences in some countries. One of the difference that I'm going to highlight is the difference between two groups of countries, South America and Central America and Mexico. South America saw a much larger decline in inequality that the countries in the Caribbean and Mexico and here when I'm saying the countries in the Caribbean please be very mind that I'm talking about a couple of countries, we don't have a lot of consistent data to construct these averages across a large number of Caribbean countries but in Central America we have data for Costa Rica, for a number of countries. The decline was slower, was smaller and the timing was also slightly different. So, of course these countries are commodity boom exporters, these countries intend to be commodity boom importers so here you can see some of the things I'm not going to be talking about the role of some of those macro shocks or the commodity super cycle. Things I've been going to be talking about is what is behind these reductions in inequality in terms of skills, here you can see one dimension of skills where inequality decline which is the returns to education or the college premium or the educational premium if you wish. There was a decline in the education premium no matter how you measure it but here I measure for college workers in two different ways, I measure the gap in wages between college and primary educated workers, I measure the gap between college and high school educated workers. Why I do this to emphasize again that there was a decline in the college premium, there was a decline in the high school premium with respect to people who have primary or less fundamentally driven by what happened to the people who were at the bottom, the ones that had primary education or less. Those wages grew much more than the wages of college educated workers. So it's not that the wages of college educated decline, some countries in Mexico apparently it did for a sub period is fundamentally the wages of people who have basic education increase quite dramatically. Perhaps less known fact but also regional fact is that the experience premium fell in pretty much any Latin American country we can measure consistently. This is the experience premium, this is the gap between people who have 30 or more years of education or people who have zero to five or the gap between people who have 20 to 30 years of education, they track each other pretty well. You can see the exhaustion in the returns to education. Sorry, so to experience, but also you can see how these returns to experience decline quite significantly. Now, one can do a hack a blinder the compositions where we bring these two elements, what happened to the two main aspects of human capital we can measure, which is education and experience. How did they contribute to the decline of inequality? And what we find in work with Manuel is that, well, fundamentally the decline in equality if you do it the way we did it just by looking at education, sex and experience as potential drivers, it was driven fundamentally by changes in the premiums associated with education and experience and gender. And fundamentally driven by the decline in the returns to education and experience, although the wage gap across women and men also declined and that contributed a little bit to the decline in the quality. In work with Sergio and Shiko, we enlarge these set of potential drivers. We look at the different period in 95, 2013 and there we find that actually the decline in the returns to experience is even outweighs the importance of the decline in the returns to education in explaining the inequality dynamics. All right, of course, when we do a hack a blinder the compositions like these ones you are basically saying, okay, there is a change that happens in inequality that has to do with changes in who is in the labor force in the composition of the labor force and changes that happen because of changes in the structure of wages. But the two things are related, right? The returns to education fall partly driven because changes in labor supply in the quantity of education in the labor market are changing. So that's what we can do in terms of asking what is the role of labor supply in the reduction of inequality. We know that there's been a big increase in educational upgrade in Latin America. Well, to what extent that increase in educational upgrade is connected to the decline in the college premium and to what extent changes in experience profiles may affect changes in the experience premium. This is what Katz and Murphy did many years ago for the US. It has been done in many other contexts. What we did is to extend the analysis for the period of a great inequality decline with Manuel and what we found is something like this. These are the residuals of basically what you can explain. So think about these regressions as follows. You are gonna estimate the production function and you are gonna estimate the degree of sustainability of different type of workers. Once you have the lastesis of substitution of these type of workers that have different levels of experience and education and so on, you can use changes in labor supply, how many workers are this kind and that kind, to try to explain changes in inequality, to relate to changes in inequality. You're gonna get a residual and that residual is what I'm gonna be plotting here. So if changes in the experience premium are completely explained by changes in the type of, in the shares of workers with different type of experience over time, this graph will be totally flat, right? These residuals, which we can call the demand for experience over time, are basically what we cannot explain with labor supply. So what will we learn from this graph? Well, we learned that for unskilled workers, the experience premium is to a great extent well explained by changes in supply. The changes in demand are very small. If anything, there is a little bit of a decline in the demand for experience over the later period. However, the experience premium among college educated workers is much less well explained by changes in labor supply. You see a big decline in the demand for experience or the residual in this case among college educated workers. So something happened with the demand for experience or the premium that the market is willing to pay for experience among college educated workers that decline over the period of the great inequality decline. Now, if we look at changes in labor supply, how do they explain changes in the education premium? We find something even a little bit more disappointing in the sense that, well, for the two countries that we have long data, which are Brazil and Chile, well, look at what happens with Brazil. The high school premium is very well explained, for instance, but changes in labor supply. But that's because fundamentally the high school premium was falling throughout these periods in 1990, 2012. In Chile instead, that follow a pattern more similar to many other Latin American countries where the premium increased during the 1990s at decline during the great inequality decline, well, labor supply trends do not catch that. What basically this residual is telling us is that the demand for skill increased in the 1990s and people have thought about trade liberalization and other potential drivers for that demand. But this demand increased, this demand pushed kind of plateaued and started to decline in the 2000s. This is true if we do college educated versus all unskilled or if we do high school versus primary, that plateau is there. But it's more important among the college educated, there is some decline in the demand for college educated workers. Now, I've talked so far for, I don't know, 10 minutes maybe, about human capital and how that contributed to the decline in inequality, but of course people have jobs and those jobs matter. One way of looking at one dimension of jobs that may matter is whether you have a formal or informal job. So what happened during this period? There was an increase in formalization. The share of workers who had a formal job in the labor market increased from less than 50% to some 55, 56% on average in Latin America. And that may have reduced inequality through a number of dimensions. One dimension in which it potentially contributed to reducing inequality is because the gap between informal and formal employees and self-employed and formal employees decline over this period, okay? But the other dimension that would basically, I don't have it here, the other dimension I don't have it here, sorry. But the other dimension is essentially because inequality for this group, the self-employed is much higher than inequality among formal workers, just the degree of inequality of incomes among self-employed. There is just a composition effect associated with this increase in formalization. People are moving from a sector that is very unequal to a sector that is less unequal. But that's not the end of the story. As with many of the dimensions that I will be looking at, inequality even within those group decline. So what this shows is inequality within group, inequality among formal salaried employees, among informal and among the self-employed. Here is the graph that you say, you see inequality among the self-employed is much higher and it declines over time. All right, I have five minutes, wow. So I'm gonna have to maybe move a little fast and skip a couple of slides. What we do is essentially, then try to understand to what extent differences in the composition of the labor market may have contributed to the decline of inequality. So what we do is a series of variance, the compositions. Think about the following regression. You have the log of wages, a set of individual characteristics and a set of job characteristics. You can make a variance, the composition. We say the variance of these wages is the variance of that job characteristics, the variance of the individual characteristics and the covariance between the two. And we do play around with different ways of defining those jobs. And we're gonna see how far we can go in explaining changes in the variance of wages just by changes in the variance associated with between group inequality across skills, between group inequality across different definitions of jobs and that covariance. If you do just skills essentially, what we find is yes, inequalities declined across skills, explained changes in the variance of wages over time, but you only explained 34% of changes in the variance of wages. 66% of the change in inequality that I am showing you happened within very detailed human capital cells. So clearly human capital contributed, but it's not the end of the story. Let's see what happens if we enlarge and we introduce characteristics of the jobs. We include sectoral occupations. We include sectoral occupation and formal status. We do obtain a greater explanation of the change in the variance. So the variance across groups of formality, sector and occupation explain 15% of the reduction in inequality. These are just the compositions. When I say explain, I don't mean in a causal sense, they are the compositions, but it's still 42% of that variance that declined that is within groups. So there's something else. Well, that's something almost maybe firm, so you can do the same thing that what I was doing before dividing the compositions, but now we look at the contribution of changes in the variance of wages of differences across workers. This is what CARD and many other people have done very recently. Changes in the composition across firms and changes in the covariance. So what you can see here is that the variance of firm effects and the variance of workers effects decline over time, but the variance of firm effects decline much more dramatically than the variance of worker effects. So yes, there was a reduction in inequality associated with changes across workers, but changes across firms seem to be even more important, at least in explaining the dynamics of inequality. This is the case of Brazil, this is Ecuador and this is Costa Rica, which is the only country where inequality in wages increase over time. And interestingly, again, it is the variance of wages across firms that is increasing more than the variance of wages across workers in the case of these countries. So I could do more, but I'm not going to because I can't, but obviously that opens the question of why is it that firms matter so much for inequality? And why is it that they matter so much for the dynamics of inequality? Well, one traditional explanation is rents, firms have rents and they share these rents with workers. So there is product market power that is shared and that's why firm matters for inequality. A more recent complimentary story has to do with the role of firms in the pressing wages through labor market power, so essentially monopsony. There is a blooming literature on the extent of monopsony power in labor markets in the developed countries, but even in developing countries, very recently, you see those references are very recent, there is a lot of debate about what extent monopsony power is important, markdowns are important in Latin America, and the conclusion of this literature, of this small literature that is emerging is that at least the degree of markdowns in Latin America are as high as in developed countries. The only exception is this paper that looks at China and India finds very small markups, but most of the other literature of the papers find markups or elasticities of labor supply that are very consistent with Latin America. But this opens, I'm gonna steal one minute. This opens, I think, very important questions that are unresolved in this literature too. So okay, we know that markdowns are important in Latin America, but to what extent they are related to dynamics that we observe, so what are the dynamics of this monopsony power? How does informality affect monopsony power? And there are two dimensions to this question is how monopsonistic are informal labor markets, but also how does it affect the fact that you have what Marx used to call a reservation margin of workers that could go into the formal sector, and that workers in principle should lower the ability of those firms to exert monopsony power or increase the ability, sorry, so you'd increase the ability of the firm to exert monopsony power on workers. Now, more important to me is the last question is is there any, what is the connection between monopsony power and inequality, and this is what we are trying to do in his recent effort with Alvaro, Garcia, and Marcela Slava. We look at data from Colombia, sorry, from Chile for the time being, but we are moving to Colombia soon. We measure monopsony power using the production-faction approach. What is beautiful about this production-faction approach, which has a lot of shortcomings, but one beauty of it is that you obtain measures of monopsony power or McDonald's per firm, so you can relate what is happening to that firm to the extent of monopsony power that the firm has. What we obtain so far is that when we separate the degrees of monopsony power across two type of workers, blue colors and white colors, those degrees of monopsony power are fairly similar but with different consequences. This is the underlying distribution of the marginal revenue product of labor for blue color and white color. You see blue colors are shifted to the left, lower marginal revenue product of labor. Interestingly, within the same firms, interestingly, the variance of wages across white-collar workers, the variance of the marginal revenue product of labor is much higher among white-collar workers. Now, when we move to wages, things are similar, but the variance of white-collar workers has declined quite interestingly. And what we find indeed is that the correlation of what we measure as markdown on wages for blue-collar workers is much lower than it is for white-collar workers. So in a way, if there is monopsony power, at least in the manufacturing data in Chile, it seems to be much more important for white-collar workers than for blue-collar workers. Again, these are not high-skill and low-skill. These are blue-collar and white-collar and we are moving now to administrative data where we are going to be able to pin down skills and repeat these exercises. But this is where we are going. Thank you, I'm sorry for.