 Thank you very much for attending this presentation and also to the organizers for inviting us here. I'm gonna present today one of the chapters of my dissertation, which I submitted last week, so I hope you don't find it too bad. There's still room for improvement, so comments are still welcome. So, well, so the paper is that policy evaluation of gender affirmative action in engineering schools. So, why should we care still about gender imbalances in higher education? In the last few decades, there's been a very significant increase in women participation in higher education. In most countries, women are either health or more than health of students who are attending higher education. Despite that, there is still a form of inequality that arises in segregation across fields of study. So, globally, only 35% of the students enrolled in science, technology, engineering, and mathematics degrees are women. These statistics gets a bit worse in some developing countries. For instance, in Chile, only 19% of the students enrolled in STEM are women. And it also gets worse when we get to further stages of the academic track. So, it is the STEM leaky pipeline. It's the phenomenon that's what we see is that women are dropping out from STEM. So, on average, only 30% of the world's researchers are women. This is a disadvantageous for women because STEM are also the degrees that are very well paid. So, there is a premium, a wage premium on these degrees, and this is one of the relevant explanations for the gender wage gap. So, more than talking about the causes of this gap, in this presentation, I want to talk about one of its potential solutions. So, affirmative action policies have been used by universities widely, in order to, one, provide better opportunities for disadvantaged groups, and this has been mainly done in terms of race and caste, and also to respond to some universities' institutional interest of having diversity in their students' bodies, and also with an argument that perhaps having diversity in the student body is good for everyone, for students who are part of the university. And the academic literature has mainly focus on race and caste policies because it's the policies that have been implemented. So, it's the policies that we have information to, and it has also mainly concerned itself with informing the first point. So, is it that these policies are effective in helping disadvantaged groups to access higher education? That means that we know very little about the effects of gender affirmative action interventions at the university level. And what would we need to know? So, first, in terms of effectiveness, can gender affirmative action policies help reduce gender imbalance in STEM? The question is relevant because maybe women are just not interested. There are no preferences to go into STEM. So, even if we provide better opportunities, they won't take them, and if they do, we also want to know to what extent, right? Like, how much can policies help? There is also a question of when we are helping disadvantaged students to enter degrees for which they might not be well-prepared to, we might also be generating a negative impact on themselves. So, if they're prepared, maybe they're gonna drop out or... And we might also be generating some negative impact on their peers. And lastly, there is the question of when the policy works, are there any further effects? So, this is about the second point of the purpose of affirmative action, of generating a more diverse student body. So, are there positive effects? Are there gains from having a more diverse student body? So, this is also a question that we should know the answer to. Luckily, in 2014, Chile was kind of a pioneer of implementing gender affirmative action policies. So, that means that I can attempt to answer these very important questions. So, the policies are implemented in the context of a centralized admission system, where two universities implemented two distinct policies, and the other 28 universities that are part of the centralized admission system didn't. So, that means that I have kind of a natural experiment setting that allows me to study effectiveness in reversing discrimination. So, what I looked at is the effects of the policies in attendance, and also on the academic ability distribution of the students. So, the part of academic ability I'm going to skip today because of time constraints, but I'm going to show you what happens with attendance, and then the second question, which I find particularly very interesting is, what's the effect of having a more diverse student body in terms of gender on students' academic performance? So, I looked at the effects of peers on grades and dropout rates. So, before I start telling you about the analysis, I want to explain to you a little bit more how the Chilean admission system works. So, the Chilean University admission system is centralized. That means that students submit just one application where they list their preferences for their, from the top, most preferred preference to their least preferred preference, and a choice is a combination of a university and a degree program. So, students decide what they want to study, if they want to be lawyers or architects or engineers before they apply to university. So, for example, a choice might be engineering at the University of Chile or architecture at the Catholic University. And the admissions depend only on a student applications course, which are measures of like students result in a standardized test and also in their high school records. But the measure is objective. It's public. And yeah, we know it. The students know it before they apply for the degrees. So, students, like the application system works with Gail and Shapley algorithm, and students are admitted to their most preferred degree for which they achieve that sufficiently high score. And because they are limited seats, not everyone gets in, but the one who gets in and the one who have the highest scores amongst the ones who apply. So about the policies now, the policies are different. The first one was implemented by University of Chile, and then the the policy increases cohort size. In the ad, so they have the regular admission system, in which they have 800 vacancies each year. Students admitting there are the top, the ones with the top scores. So the proportion of women and men only depend on applicants. But what they did is that they added 40 extra female exclusive vacancies from which they draw from the first 40 women in the student, in the, sorry, in the waiting list, and they accept all of them. In this way, they increase, I mean, they are aiming to increase the women participation in the engineering school by having, like, it would have at least a mechanical effect of four percent. The second policy is from the Catholic University book, and it's a bit more complex. It's a bit more nuanced. This policy, what they did was they tried to address women's preferences, and they changed a little bit their curriculum, and they increased the combined engineering with disciplines of more interest to women, such as biomedicine, architecture, or design. So it's a, it's some policy that's more complex and probably a bit more costly. And they also increased the number of female faculty members. So then, what's the effects of these two policies on on applications and attendance? So when looking at applications, I'm gonna look at that, I have the admission system data. I'm gonna look at all the students who applied to engineering at Uchile or engineer at a PUC in the first choice of the rank or in the second choice of the rank. And then I also looked at who are the ones who are finally admitted. What's the proportion of women who are finally admitted in each of the schools? And in the case of the University of Chile, I'm interested in the total proportion of women admitted, so that women admitted through the regular admission process, plus the ones admitted through the 40 female exclusive seats. But I'm particularly interested in the first part, the regular admission part, which I believe that I hypothesized that it could also be changed. And how could it be changed? Well, if the policy changes women's preferences. And this might happen if, for instance, women have preferences for studying at places with more women. So they see, oh, there is a policy that's gonna be increasing the proportion of women at its engineering school, then I'm more interested to apply. And it's not so crazy because there is a long literature of how women benefit, have better results when they are with more women in the environment. And the other option is that there might be changes in aspirations. So the University is saying we have this policy to attract more women, and women might say, oh, well, maybe engineering is a place for me. So it's something that I start considering because of the policy. Even when I had the sufficiently high score anyway. So to measure the size of the fact, I've used a difference in difference approach with two treatments, one for each school, because the policies are quite different. And I used the 28 other engineering schools as a control. And I estimate a difference in difference equation using university fix effects and then fix effects and the two treatments. So what do I find these first results for applications is that there is a 10% increase in the proportion of females that are applying to the University of Chile as their first option and a 5% increase in the proportion of females applying to the Catholic University in their first option. So both universities increase the applications scores by a significant percentage. And then in terms of attendance, we also see at the University of Chile, so the first column, this column shows the results on regular admissions. So the part that shouldn't necessarily be impacted at the University of Chile by the policy. And we see that there is a 4.4% increase in regular admissions. And the total effect, including the 40 female exclusive seats, it's a 0.4%. So this means that the hypothesis is true, that there are more women applying even when they meet the criteria without the lowering in admission criteria. So there are more women interested in attending engineering school even when they have a sufficiently high score. And the increase at the Catholic University is of 7.2%. So it's a bit lower, but the difference is not significant. So both policies had very strong, strong effects in increasing attendance to the engineering schools and these effects are significant and similar size. Then the second part that I'm interested in is to see, well, we saw that the policy had an effect, that the policy had the effect of increasing female participation in engineering schools. So now engineering schools are more gender diverse. What happens then? What happens to the incoming cohorts of students where there are more women? So I use that from the University of Chile to determine the effects of gender composition across peers on male and female engineering students. And I looked at grades on first year subjects, grades on a collaborative project, and also at dropout rates. So why do I focus on the University of Chile? Well, first, the policy is a bit more controversial. So there might be more opposition to it, but also because the University of Chile provides with an excellent setting to study peer effects at university. Studying peer effects at university is difficult because the students tend to self-select into their courses. They can choose what classes to take and they might choose to take it with their friends. So it's very complicated to estimate peer effects in university. And, well, what happens in University of Chile? There is no self-selection. What happens is that before engineering students set up food in the school, university administrators have already allocated them to groups, which I call classrooms. So they are exogenously assigned to classrooms of around 100 students each. And that means that there is exogenous variation on the gender composition across classrooms. I have around 71 of these groups because I have several cohorts. And, yeah, I do find that there is some variation in gender composition. The other thing that's good about this setting is that I actually have data on the relevant peers of the students because these students are allocated to these groups. And in these groups, they have all of their teaching activities together. So lectures, seminars, any kind of teaching activity. And it all happens in the same classroom, so in the same physical classroom. So maybe I should say here, this is my alma mater. I attended this school. All of my friends in the first term were in my same classroom. We didn't have time to see other people in the school. We were all there together the whole day. So this means that I can estimate the gender peer effects using a simple linear in means model. So this is the model where the outcome is either grades or dropout rate or dropout. And this outcome is explained by the students' own characteristics, including measures of academic ability, some of their peers' characteristics, but the outcomes of interest is the proportion of women in their group. So what do I find? Very quickly with grades, I find that there is no effect of the gender composition on the core subjects. So this is physics, calculus, algebra, computer science. There is no impact on a student's academic achievement, neither for men nor for women. But when I looked at the collaborative work project, I find positive results of having more women in the group. So interactions are a bit more intense in collaborative projects, and I find that there are positive effects there. But what's perhaps the most important outcome here is what happens with dropouts. So I look at the probability of dropping out after the first term, and I derive a binary variable called dropout, which is one if the student doesn't enroll in any courses in the second term. So this part I think it's very important because it's relevant to understand the reforms effects on persistence in the engineering school. So I mentioned before the existing leaky pipeline in STEM. This outcome is relevant to better understand that. So I run the same equation, the linear mean model, but it's a logistic regression because the binary variable. And I find that there is actually a reduction in women's probability of dropping out when there is an increase in women percentage, and to translate this into something a bit more meaningful, I'm going to show you the predicted probabilities of dropping out. So before the reform class, the groups were around 20% women in each classroom, and at this level we have dropping out rates of around 3.8% for women. So the reform is creating an increase, an average increase of around 8.8% of women in the classroom. So to better understand what would be the effect of the reform, what is the effect of the reform, we looked at what's the predicted probability when the percentage of women is more around 28%, which is around 1.4%. So that means that there is a drop in the probability of dropping out of around 60%. That means that the reform can help to increase the persistent of women in the program. So just to sum up, in this paper I take advantage of natural experiments and use a difference approach to investigate the effects of gender affirmative action. I find that the policies are both effective in increasing female attendance at the university, at the engineering schools, and not only that, the reform also has effects in increasing persistent at the engineering schools.