 So good afternoon. I'll present joint work with Rodrigo Oliveira and that's on Severnini. Rodrigo Oliveira is from UNO wider and that's some Severnini is from Carnegie Mellon University and this paper is it started as my master dissertation and it changed a lot and it's about basically affirmative action in a top university in Brazil with no major switching. So basically what we are going to try to answer is when you implement affirmative action in a university what happens with the students during university and then it's not in the original paper but we are doing a new paper now and we try to estimate the impacts on labor outcomes as well. So this is the presentation framework with we start with the institutional background and data then go through the empirical strategy show some academic outcomes and mechanism labor market outcomes go through some robust neck checks if I had time and then do some concluding remarks. So the motivation is in 2014 one 25 percent of all universities had of all countries had some sort of affirmative action okay so countries like United States Brazil, India and Israel had some kind of affirmative action for higher education and we have we have we have papers for the United States that show that in their settings if a student goes into a major that he thinks he can't finish he can change majors but this is a lot more difficult in Brazil because in Brazil we have a unified exam to enter higher education and if you don't you think you can't finish the major you have to take that exam once again and it's a really hard exam so it's harder to switch majors. In Brazil we have little evidence on how they adjust so they can change majors so how they adjust to not being not when they enter if they're not able to to to follow their the other students what do they do if these policies target the people they want to target so the idea of this policy that we will call the quota policy is to increase low income students into higher education. The magnitude of those effects if there is mismatch so all the students that are getting into higher education or they're prepared to to finish it and finally what are the labor market outcomes that is a new thing that I'm working on so the labor market outcomes is a fairly preliminary work okay so again we study framework where we have no major switching and rigid curriculum so 90% of the courses in higher education in Brazil are mandatory so I don't have that kind of I can't switch to easier courses because I have to do 90% of the courses then again we study the time period between 2003 and 2006 and I'll explain afterwards why we limit our analysis to this period okay so basically no manipulation of the quota policy and I'll explain it explain why limiting the study to these years gives us the impression that it would be harder to manipulate the quota policy and the main findings is quota students do have lower background they and the quota policy does not increase dropout significantly it's fairly similar to the US and framework where you have you can switch majors so if you cannot switch majors and you have the same dropout it seems that that's not such a bad thing quota students catch up during university the ones that graduate so they start worse off and they catch up to their non quota partners and then quota students use graduate hours to adjust in the beginning of the course and finally that quota students that only entered higher education because of the quota policy have lower wages than their counterparts okay and then we we show some heterogeneity by fields so the existing literature for for the US we have some older on papers by car by long cargo Dixon and bleemer that mostly use the proposition 20 to 209 from the US to estimate the impacts of affirmative action but aside from bleemer they did at the identification strategy is not so good because it's hard to know if you get or not into higher education is more subjective than in the Brazilian setting we have the papers from Bertrand and Badge for for India where the setting is really similar to the setting in Brazil and then for Brazil we have some papers for for when you bear from the Francista Antony Pianto he better fallage is seven for when you come to salon which is here for nationwide and mail for for the university of the spirit son thanks and then we have some evidence on mismatch and dropout for the US and for major switching and what is our main contribution we should show that mismatch is relatively low in our setting okay so the students that get into higher University they they have lower background they have lower scores but they do graduate okay also we show the mechanisms of adjustment for the for these students to graduate we use a previous ability control so we control for for the entrance score and we show that in some settings the the quota students go better than their counterparts and finally we show inputs on labor market outcomes that only one paper by Tony Pianto for one bit did so we show it in another setting and also we use a different identification strategy so this work is divided in three parts now because we have later labor market outcomes and basic basically we measure the effects of academic outcomes we show the mechanisms and we measure the facts on labor markets using former employment and wages so the Federal University of Bahia is tuition-free okay the you you have once a year an exam you had in our time period once a year an exam called the Shibula that you did and you you chose the major beforehand so you you you chose the major before entering University we have a rigid curriculum the most prestigious university in the state of Bahia and it's located in in a state where eighty five percent of the the people are black are mixed are mixed rates are black okay so the quota system was announced in 2004 and implemented in 2005 and it reserved 45 percent to low-income students how did they target low income low income students they targeted public high school students students who had done all their high school in public schools in one year of middle school in public schools also as we had 85 percent of the population black we have 85 percent of the slots reserved for black students okay so we have three administrative records linked by an identifiers so we have student registry at UFBA where we know the academic the academic outcomes for these students we have the social economic questioner where we have some some characteristics for the students that we use to control in our regressions and we have the Ministry of Economic Labor Market data where we have employment and wages for the formal sector so this basically is the vestibular score before and after the quota implementation so what we see here is this solid line here is the quota students score vestibular score before affirmative action the red line is the score for students who would have entered even if the quota wasn't in place the blue line is the score for students who only entered UFBA because of the policy so we show that the policy there is a shift to the left for the students on the distribution of score and for non quota students we have a difference only for the students on the margin of entering UFBA okay and when you look at GPA final GPA we see no evidence of differences for non quota students before and after or for students that would enter in the same way so the difference in GPA comes from the from the group that only entered UFBA because of the policy empirical strategy will be our control group will be basically the 55% best-ranked students from private schools I have a picture in the next slide so it's easier to see and the treatment is all students who attended public schools okay so basically if I have a major where I have a hundred slots and before the quota policy I had 80 students from private schools and 20 students from public schools after the quota I would have 55 students from private schools 20 students who would enter the same way and 25 students who only entered because of the policy so what we're going to do is we're going to compare the 55 best students from before with all the students that entered after so that we don't have the selection effect for a control group and for the the treatment group we have two effects we have the effect on the group that would enter anyway so that we will see that there is no effect basically and then we have a selection effect that is due to the students that only entered UFBA because of the quota policy my treatment so my treatment I'll have a dummy for for quota students there are those 20 before and those the 45 after so okay so basically I estimate two main equations first always I'll have a dummy only for quota students and then I'll have a dummy for quota students that would have entered and quota students that would not have entered without the policies so I can I can do an extra genius analysis so so basically this is the result for all students that enrolled at enrolled at UFBA so we see an increase in failures okay a decrease in graduation rates at 12% decrease in graduation rates so the first two columns we don't control for the Sibler score the last two columns we control for the Sibler score but both for failures and for graduation we see that the fact comes only from the group that only entered UFBA because of the policy and when we control for score the fact on failures vanishes and the fact on graduation still it's lower in magnitude but it's still there so maybe there's something else besides the scores that that predicts graduation rates then when we do we do a third genius analysis for the areas and for the technology majors what we find is the graduation effect is a lot higher so we have 20% decrease in graduation rates that are still there if we control for scores and they all come from the the group that would not have entered if the quota policy isn't there so the technology majors are the majors that are driving the effect on graduation then when we limit our sample only for students who graduated we see that the failures we have an increase in failures a 30% increase the same thing the first semester GPA we have a decrease but by the end of the course that by the end of the major it has its half in the intensity so there is some evidence of catch-up and when we control for previous ability we have a positive effect so people who come from public schools that have the same score as people who come from private schools do better in university and maybe it's because they have some other skills that maybe they they are like they have more ability and they had they had a poor high school education so they catch up and they even overcome their their private school counterparts and this is true for both the groups okay all the effect without controlling for ability comes from the the students who only entered because of the quota policy but the effect that they do better when you control for score is valid for both groups okay so this is basically the summary of the results that I just just told you okay and when we go semester by semester this is what happened with average grades semester by semester in blue we have so this one is change yes so the blue one usually is the group that entered would have entered without the quota policy and the red group is the group that only entered because of the quota policy so we see that in the first semester we have lower GPA then it goes up it goes down again and then it go up constantly until the end of the major and this is the great average including all semesters before the first one so this estimate is equal to this one and this one is different from this one because it is a weighted average of these two estimates and we see that the catch up is constant along the major okay failed courses it's basically the same thing they fail more courses in the first semester then less then more again and then it goes down so what it shows us is that students are catching up and then these nest graphs show that quota students they take the same amount of courses in the first semester but then they adjust they take take less credit hours in the second third and fourth semesters of the curse and then they go up again ending with more credit hours than their private school counterparts this is made mostly due to retaken courses okay they take it they fail the curse and they then they take it again for labor market outcomes basically it's the same thing true first columns we don't control for score to less columns we control for previous ability so we see that we don't have an impact in employment so informal employment but we do have an impact in log wages so quota students after the quota policy are earning nine lot lock points less than before also once again defect comes only from students who would not have been admitted if the quota policy wasn't in place and for these students it's marginal but we see an increase in the probability of employment with when we compare comparable students according to score this is the counterfactual is me if I would have entered Ufba without without need of the quota policy so no okay only for the last two columns yeah I have okay there's a type in the last line yeah and then when I go only to the students who graduated I have no effects and when I I do the detergent naysay the heterogeneous analysis I have a positive effect when I control for scores for students who only entered because of the quota policy both on employment and on log wages okay so because they have lower scores in general and they you don't see an effect on log wages but they have lower scores and they have the same wage so you have a positive effect on those wages when you control for scores and now it's the same thing as before but now it's years after they entered Ufba so this is for employment and employment control for score so what it sees that those students and this is for our students so for these students the quota students I see a decrease in the amount of students who were in the formal labor market when they enter Ufba so maybe the quota policy has some effect of saying no I have this opportunity now I want work from the beginning but in the end we see no effect and on wages we see no significant effect maybe it's a problem of power here and I'll do the heterogeneous analysis in the next slide so we can see and when we control by wages we see no effect and this is the heterogeneous analysis for students who would have entered and wouldn't have entered so we see that we have an effect for the students had that only entered Ufba because of the quota policy and when we control by scores this effect vanishes so we do a lot of robustness checks that I'll not go through them now because I have no time but the concluding remarks is basically that the quota policy had a positive effect on failures and a negative effect on graduation rates but those effects on graduation rates are not so high when we compare two other settings among college graduates you see the same thing but you have some evidence of catch up when you control for ability most of the differences disappear it's okay and basically students are just credit hours okay when to to be able to graduate and finally students who only entered Ufba because of the quota policy have a lower wage when you compare them with quota students that would have entered Ufba even without the policy and that's it thank you