 I'm Gabriela Zapata, and I'm going to present the role of the family of origin in shaping inequality opportunity in Chile. The red one? Sorry. Well this paper investigates this inequality opportunity measures using data from Chile. Red summons the evolution of inequality opportunity in 11 years period from 2006 and 2017. With different circumstances, one is parental background, family composition, gender, region of birth, ethnicity. And what is the role of these circumstances is shaping inequality opportunities in Chile. So in particular, to what extent, circumstance at birth determines or are associated with labor market outcomes. So this is grounded in inequality opportunity, Chico explained really well all the theories, so I'm not going to spend a lot of time doing this. But I use the extant approach as Vito, and analyzing secondary data from two waves of the survey Cacen. So after Vito's presentation, probably I'm going to expand it and I'm going to use all the years because I already use it in different papers. What the ethical justification is basically that not all sort of inequalities are the same, and that you can sort of split inequality in the one that is unfair or that comes from circumstances that you don't have control over with. And the part that is not, so it depends on your own effort. So and this theory, sort of this methodology basically split these courses and try to identify the circumstance one, which is the extant approach. So basically in a world with equal opportunities, the play field is level before the race of life starts to compensate for uneven circumstances. Over where individuals don't have any control, like I always exemplify with this playing field, where you have bumps and the one that is already like leveled. And as Vito mentioned, it's a very attractive concept because it holds individuals responsible for their actions and choices. So what is the country context that the motivation, well Chile is a country in South America with a very high inequality among the highest of the OECD countries, with a Dini of 0.45 in 2008. We have a very high concentration of income at the very top, with a 10% concentrating 72, almost 73% of total income, while the average in Latin America is 54. The top 1% concentrates almost 30% of total income. Also we have a very high intergenerational persistence in education income. So with income elasticities of 0.6, correlation coefficients in education of 0.4, and this absolute mobility is combined with very relative persistence at the top of the income distribution. So we also have higher returns to higher education that has been dropping over the years. Also with a great dispersion depending on the quality or prestige of the higher education institution you attended. And the background of the family. So basically what I said before, we decompose inequality in inequality opportunities and the rest is effort, inequality, plus lack, plus everything that we don't measure. As effort is a private information, so we attempt to measure differences in outcome, do it to different circumstances, holding effort constant. And the means among these circumstances is the measure of inequality opportunity. So in practice what we do is we apply the inequality index to a counterfactual income distribution where we eliminate all the inequality that doesn't come from circumstances. So Tic also used this kind of like something like this figure where if our sample is divided for example in four circumstances, so this is parental education. So we have people with parents with no formal education, parents with completed primary education, secondary and higher education. So basically the way that we construct our counterfactual distribution is by averaging income of every type that is called in this literature. And then we compare. So with this counterfactual distribution we apply an inequality index and that's the measure of inequality opportunity in the ex ante approach. So I'm not going to explain the parametric method because Tico and Vito also mentioned them but basically what you do is the same but parametrically. And then you can have the level of inequality opportunity which will be the inequality index applied to this counterfactual distribution or you can express it as a share of total inequality. So comparing the inequality in this counterfactual distribution with the total inequality of the sample. Then the next question is which inequality index in the beginning this literature to use the mean logarithmic deviation because of the decomposability property. So basically because you are sort of splitting total inequality in these two components efforts and circumstances then the natural inequality index is one that is decomposable so it was using mean logarithmic deviation. Here the MLD is more sensitive to a lower tail distribution and because the way we are constructing our counterfactual distribution is by average so we are sort of eliminating the tails. So the latest Vito was part of the paper said we should use a genie index but the problem with the genie index is that it's not decomposable. So the genie index cannot be decomposing these two parts. So to solve this problem in this paper I use the Shapley decomposition which allows the estimation of the marginal contribution of the within and the between groups because basically this is what we are doing here. So the inequality of circumstances is the inequality between groups and basically the one that we are removing the inequality of efforts is within groups inequality. So the way to do it is basically by averaging this. If we apply the genie index first of a counterfactual distribution done by this average by types we get a different result. We are assuming that the rest is inequality of efforts or effort plus lack plus everything else but if we do the same but of the other way around so if we calculate our counterfactual distribution putting the same average income by types then we will get a different genie. So this method sort of averaged these two things and we will see a kind of like in between estimate. So it's higher than the MLD that it was criticized in the beginning that the MLD gives lower estimates because it's focused on the or it's more sensitive to the lower tail of the distribution but the genie is better because it focused on the middle of the distribution but here we have something in between. So the data comes from as I said two waves of the Cassin survey. We use a sample that is a household heads at working age men and women 25 and between 25 and 60 who are active in the labor market. I have two outcome variables, individual market income and individual hourly incomes. The idea is to try to correct for the women's selection. Women's were not included in the first papers because it said that they sort of work less hours and that is an endogenous decision so I use hourly income to solve for that problem and the circumstances is parental education, gender, region of birth, ethnicity and family composition by family composition is if the person lived with both parents until the age of 15. So this is the sample. We have more women on the second year. Parental education, increase in primary and secondary and higher education the second year. I have a problem that I have to solve because it doesn't make much sense that people without any education or primary and complete increase in 2017 and that has to do with the way that the question was asked in 2006 so maybe it will have to use 2009 as the first year to be comparable. Then the place of birth is divided in four regions if you know Chile is very long. So basically climates and industries are very different between the regions. So it's the north, the center, the south and the capital which is a metropolitan city and this is the percentage of people growing with parents. So it decreased a little bit between these two years and people with indigenous background. So what I have here in the first descriptive slide so you can see in a world with equal opportunities I think it's easier to see it and I don't have the other graph. So this is average earnings in Chile and Pesos for different subsamples. So this is 2006 and 2017 so if the income, the average income doesn't depend on your circumstances you wouldn't have a reason to see this slow. So you will be flat, average should be equal though they wouldn't have a reason to don't be different. So what you see here is actually a very high correlation between your income and the level of the education of your parents. So people with parents who don't know education are in this blue color and you see the difference with an average team of people that have parents that attended higher education. So you see the same in both years, you see higher average in the second year, you see higher average for men than women but it's very clear that your income depends or at least is highly correlated with the education of your parents. And here are the measures, these are the measures of inequality opportunity. I defined four scenarios so basically to compare with other countries. A scenario B, A is considered as a circumstance only parental education and that's the estimates of inequality. This is the genie of the actual distribution. This is the genie applied to the counterfactual distribution but in the way that I showed in the previous slide and this would be the within and in this case we can add both and it will sum exactly the genie. And then we get almost 30% in 2006, a little bit less in 2017. So it's not a big difference between the two years. Then when we add a circumstance so in scenario B what I have is I include gender so it's parental education plus gender and then scenario C I add region of birth and in scenario D I add family composition and ethnicity. So this is of course normal that inequality opportunity increase where estimates increase when we add new circumstances and what we see between years is a small decline but very, very small. So in 11 years inequality opportunity has almost not changed in the country. So this is for market income, this is for hourly incomes. So basically with hourly incomes we get smaller estimates which is kind of what is expected because using hourly incomes the differences between for example genders are smaller and also because it's a household's head and women usually are in charge of the caregivers responsibilities in the house so it's more likely that a man have more than one job than a woman and things like that. So this is the sharply the composition the same that Vito showed but in a different way presented in a different way where I can see, sorry, we can see that the most relevant circumstance is parental education. For in the case of market income the second and increase the relevance so it's gender so and also for the, sorry both says market income but this one is hourly income is the other way around. No, this is hourly income, this is a mistake in the slide. So gender is more relevant on the second year in 2017. Region of birth in this case is the third circumstance relevant but it decreases relevance between those years and we see a small contribution of indigenous background and smaller even for growing with the parents. So this changed a bit when with these two income sources but follows sort of like the same pattern what is expected also that gender is less relevant in hourly income so because in hourly income we are correcting for this difference hours that women can work. So and this is the comparison between the estimates using the MLD so this is the inequality opportunity that we would have get using the mean logarithm deviation. This is the inequality opportunity that we would have get when we use the DINI applied directly to the counterfactual distribution and then as a percentage of total inequality and these are my estimate so they follow the same trend as the MLD but and they are slightly higher so the argument would be that applying the DINI index directly to the distribution to the counterfactual distribution could overestimate, could overestimate the inequality opportunity. So something that I just did in not very long ago so it's a new thing if I have comments how to interpret this it could help me too is a riff the composition. So this basically is the composition is an adaptation of the Osaka Blinder the composition which allows you to measure to identify this two composition and an instructor effect in different part of the for different distributional measures not just the mean as the Osaka Blinder so in this case I use the quantiles so and this is measuring the difference between two years so between 2006 and 2017 and how the different circumstances affect this this to the the variation between in income between these years. So we have the different person tiles here we have the composition effect measure the changes due to the varying the circumstances how this for example there are more women in the in the in 2017 in the sample people have more higher parents have higher education and secondary education more than in the previous sample etc so it's basically the composition of the circumstances and this in the earning structure effect we have the changes in the returns of the circumstances so what is dominating here is the earning structure effect and it has a positive effect of earnings so we increase earnings and increase earnings particularly on the top of the distribution just to finish this is my last slide. So the sad part is that circumstances which I kind of expected because this model is not used in this type of literature yet so it uses characteristics are for example what yender is one of the characteristics but also years of education of your own education here I'm using parental education so there is more like an indirect effect even in the changes in income in earnings between these two years so there is a lot that is not explained by the circumstances. So this also happened when when you use this in labor market settings so with other type of characteristics but what we can see here is inequality decreasing kind of effect of parental education gender is has a positive effect in increasing income particularly in the beginning in this lower part of the distribution and the opposite is for region of birth so that's it thank you.