 Thank you. My name is Sergio. I'm going to present this paper that is about Whether it matters where you were born in terms of educational mobility So I guess a good motivation is the previous presentation where you see that as a sequence region matters a lot So in my paper is an attempt to try to get some type of causal estimate of the effect of regions on education mobility in Latin America Just to clarify the the kind of the concept of mobility them that the paper focus on So we have right to we have been talking on this conference about the concepts different concepts So we have the absolute mobility concept where it's basically the progress in absolute terms relative to parents and Relative concept where it's like where are you with respect to your peers? It hold that relate to your parents where we respect their peers in this paper I focus on the absolute mobility concept and I'm going to measure that and ask the probability of children With parents That didn't complete primary education What is the probability of them completing primary education? So that kind of moving up in terms of educational entitlement I also have in the paper some results using secondary education that might be kind of a margin That is more relevant for the current cohort But I'm going to be using senses that are kind of old So I'm going to show you that the primary education here is an important margin So just to motivate you thought to start if you compute Upward mobility by region in Latin America And you map those estimates you can see that within countries you have a lot of heterogeneity so For example, you can see in the case of Mexico in the case of Brazil You have regions where the levels of our mobility are very different to other regions So then if you take these results right as given then the question or one question that you might ask is Why and is there something in the region that causes these differences? Or it's just that people are different and they choose to live in different places and that's why you observe this I'm sorry in between the two plots Sorry, just more this is kind of level of Provinces and that's level of district level So they are this basically the same just more finer on the on the right But that is the same so it's the same data It's just that here the regions are kind of more aggregated than in there in the in the plot on the right And of course if you compute this at the level of country Then you're gonna miss all these heterogeneity within the countries and of course if you think about a level of a time and especially in Latin America where you have seen an important kind of mandatory schooling Then why if there is mandatory schooling then within countries you can see all this variation, right? So we didn't there is something Within countries happening So then as I was saying then there if there is this heterogeneity then the question is is because Absorbing or is because there is something in the region that might be peers some policies on Geographical factors that explain these differences And this tincture is important in terms of public policy It might be important because implies different policies like moving people or just focusing resources on some particular regions So What I'm going to do is I'm going to exploit differences in the timing of children moves Across provinces within countries to isolate the What is in the literature called regional? Childhood exposure effects. So it's a place effect that depends on the age in which you move So in the paper you contribute to the territory by starting this place effects in a new setting in Latin America I'm going to be replicating the approach that was Applying Africa by a lecina, which is the same approach or is based on the shettian Henry paper That uses movers in the US across commuting zones to identify these effects So sub review findings. I find evidence of childhood exposure effects as well as significant sorting So they start in of course people choose places and they move to better places And what I estimate is a convergence rate of three point five percent The barrier of exposure between ages one to eleven which implies that if you move to a place with Higher level of upward mobility you catch up like three five percent of the difference is what you get in your chances of Moving up and I also find some selection effects that are around 42 percent and are very similar to what is Is has been found in other play places where this has been done And I also document childhood exposure effects using secondary education, but I'm not going to show those results here But basically you find patterns that are very similar So let me talk about the data. So I use for this paper 21 census that I obtained from iPods international They spawn 11 countries as you can see in the years They have They I don't have the same number of censuses for all the countries and they are relatively old. So only a few countries have Sensuses that are in the 2010s and then the There are more of course sensors available in Latin America But these are the ones where you can identify through the questioners who moves like you have the region the big location And you have also how long have they be living in the current place So so I bones reports these two levels, which are the ones that were in the map Provinces that are kind of course a military unit similar to the states in the US So for example in Brazil are states and districts are kind of finer and minister the units and similar to counties tuition in the US Then there is a variable reporting the province of the previous residents and another reporting build place And the number of years living in the current place So with that you can identify or classify people between movers and non-movers so people who live in the same place where they were born non-movers and then people who Leave in a different place. They are the ones that basically immigrated within the country In terms of education there are two variables And I'm going to be using the categorical one which is Has four categories Copy a lesson primary complete primary complete secondary and computer sherry This one doesn't reflect any particular system educational system in the in the continent is more Try to follow this is harmonized in a way to try to follow this standard of six three three life six primary lower secondary three and upper secondary three years And I also created a variable containing average parents education use you the provable father Our mother identified by a poems using the relationships to the head of the household. So I'm using co-residents here Corresing samples So if you kind of do kind of first descriptive plot just to show that What I was mentioning before that in the sample given the time of the censuses if you look at the individuals in the sample age 14 to 25 You can see in this axis the education level of Or the fraction by attainment of the parents and then on this axis the likelihood of child attainment And you can see that lesson primary is a relevant margin for this particular sample population and then Also, you can see that if you take the ones that with parents that did not complete primary Then you're very likely to also not complete primary. So you have cannot have Yeah If you move in if you instead of doing that you take kind of older individuals That coincide of course less with the parents And you use You map them by secondary education you can see that The level the share of population that did not complete other parents that the no complete second that is much larger and then The chair the complete secondary of children are so is much smaller Okay, so now I'm going to go to the intuition of the empirical strategy So this consists of the following suppose we have two regions a and b and they have different levels of our mobility measure as probability of finishing primary for kids with parents that did not finish primary and Ideally which you will take right randomly assign people to do to different regions and see how they do but Of course are not feasible. So what what I do here is to estimate the level of mobility using non-movers as a prediction of the level of mobility of those who move And see where the where they how they do so then In the plot These are the two regions one region a with low-level of mobility region B with High-level of mobility and then you can map individuals or compute the average Probability to finish a primary by H and move a map in the plot and you can see if They fully converge that you're going to see that they are going to be close to to the region B Right the destination so they're moving from a to B If you become like the region of destination that you should be kind of near the red line here But if you don't converge to the region of the nation then you should be around here And if you see this pattern it means that Basically the latest you move the less you converge right and then the Selection effects can be identified in this region Because we're talking about primaries then the idea here is that if you move later in life And you didn't complete primary then probably I mean of course we know that some people still Later in life complete primary but but should be kind of a no effect there and This was originally done with income so which is Better because you could for example measure income at the age of 30 years old And then you move at the age of 32 of course that should not affect your income at earlier, right? So then as a prediction of the level of mobility of the individuals I'm going to compute The probability of finishing primary for children who did not with parents who? From children with parents who did not finish For using non-movers so that's going to be the main variable and then the variable of interest is going to be this Delta ODB which is a different the mobility of the number of the non-movers between for the cohort Be between the destination and the origin So then for somebody moving from a to be this delta is going to be the mobility of the non-movers in the region B minus region a And then the specification is just this one where the dependent variable the dummy variable where an individual completes primary The sample is composed by individuals with parents that did not complete primary and then you have some fixed effects by the age of move So fix effects by the region of origin and cohort household fixed effects if you want to compare only ceilings and The main coefficient of interest here is going to be this beta That is one beta for each age of at move That interacts with the delta ODB. So it's how the delta between the origin and destination Translates into your achievement in terms of education. So those those 20 betas are basically the ones that were in the plot, right? so That's the semi parametric approach So when you estimate that you're going to have this beta that captures the causal effect But also selection because people don't move randomly right they choose where to move So this is going to be basically the two components and the key assumption for the identification here is That the selection effects don't change With age at move so basically that people who move with kids when they were to they are Not systematically different to people who move when the kids were three for example If you buy that assumption then you can separate the effect of the causal effect from the selection effect Basically doing the abstraction in the area where there should be no causal effect because you are moving later than primary So I'm going to just show you the the main baseline results First this is an histogram of the deltas. So What is the typical difference in terms of mobility for the movers and you can see that on average people move to places Where there is more I've got mobility, but but still you have People who move to places with lower Our mobility And then this is the main plot basically Where you have the h and move and then you have the dots and just align kind of Based on relevant kind of type of ages. So if you move before the age of schooling, there is some kind of flat area Then there is this slope which means that the if you move in this by adults ages the effect kind of goes down and then The area where you can identify the selection effect which is when you move Later than the school in age relevant for primary, right? So if you move at the age of 18, then if you move to a place with higher mobility that should not affect your primary completion so then Here the average is low if you take all these dots until the age 11 is 3.5 percent and Then the level of selection here is the 42 percent in terms of interpretation of this is that the convergence rate is Of 3.5 percent Per year of exposure, which means that if a children move at the age of one will pick up about 35 percent of the difference between the origin and destination on average, right? Then this rate of convergence Is just a little bit smaller than the four percent Found in the u.s. For income mobility across community sons and Higer than one that has been documented in in Africa in the case of Africa with education the same metric in the case of The u.s. Is income mobility so it's very different But they all kind of the plots of the same approach capacity looks very similar. So that's very Kind of reassuring and then the selection effect is 42 percent Which means the families who move to a region where permanent residents have for example turn percentage points Higer chances of completing at least primary. They already have four point two percent More chances themselves. So that's the magnitude of the selection that people who move they were already They had already more chances of having a mobility Then I have a set of Validation results are relatively standard in this Literature that is not big, but there are several papers I'm just going to show you the house will fix effects anything I'm going to just go quickly over the other Instrumental variable approach that I have in the paper because I don't have time but here is only that and in the first In the main in the basic in the main results. I'm comparing Families who move doesn't matter who what family but then in the here I'm adding the family fix effects of the house will fix effects. So basically then the comparison is Between siblings so a family moves. They have two kids different ages. They moved to a different place So what the one of the kids is going to be exposed to the new region more than the other? so then That's the idea in the sample of household fix effects are here and writing the same regression as before but with With the same with the same sample as here because of course when you add these household fix effects Your the sample gets reduced because some families have only one child. So they are not going to be in the sample and you can see that this basically is not something about the sample and You can see here that the selection basically goes to zero Then the slope is relatively similar and you have the same shape just kind of going down Then the other things Given this shape another approach to estimate this will be to fit a parametric Right kind of kind of these lines just assuming that the the all the deltas here the same Kind of to just estimate this is slow. This is love and this is love This is what is here, but not going to do it is just to show you that if you do that dimension You're going to get kind of those deltas only and I do that because then I kind of address them the genetic Using I'm here. I'm going to ski today to just show you that I use some particular flows of periods where there is some Anomalous migration of flows so I take all the origins. I Can't know how many people are going out then I feed kind of a trend And I take the residuals around them and then I I go kind of restricting the Sample to flows or periods where more people were going out so it's They are not they are likely to be Be pushed out of the region more than they decided to move because of the educational reasons When you do that, basically you kind of stay in similar levels in terms of the slopes Then the other concern is that people and I just only decide where to move so then here. I just follow kind of the typical Chief chair approach where you instrument the the destinations Here is the first stage then there is the the It's not the first stage them I Forgot the word now, but anyway in the in these columns You have the reduced forms. Sorry. This is the reduced form and here you have the two stateless squares And the results are consistent basically when you instrument the the destinations with these predictions using previous Migration and Would you mix the two approaches same similar patterns? So so that that's it So basically what I find here for Latin America is that people who move to two places with higher mobility they kind of catch up with the To the levels of mobility of the destination Which suggests that there is something in the places? And that is not just sort in In this in this variation in in our mobility. Thank you