 This paper, Lucky Women in Unlucky Cohort, this is a joint work with Ines Berniel, Leo Asparini and Mariana Marchioni. Well the motivation for this paper is that we have extensive evidence from developed countries showing that when workers enter the labor market in a bad time when the unemployment rate is high, they end up being penalized. They face lower earnings and worse career prospects in comparison to workers who enter when the unemployment rate was lower. But these results not necessarily apply to developing countries where the adjustment to negative shocks might be different. For instance, because female labor force participation is low and many women act as secondary workers. They enter the labor markets during crises and where there are men in the household who lose their jobs. And this change in behavior or this entry to the labor market during crises may include young women who are doing their school-to-work transition and who may find in the negative shock an additional incentive to enter the labor market and help their families. And this unexpected entry to the labor market may end up having long lasting consequences in the labor market attachment of these young women and their earnings perspectives, for instance through changes in human capital, through changes in empowerment, in perceptions. So what we say is that this negative shock might end up being beneficial for these lucky women in unlucky cohorts. So what we are going to do in this paper is to analyze gender differences in the effects of facing adverse conditions at the moment of labor market entry in Latin America, which is a developing region and very unstable region where female labor force participation is low and where we have extensive evidence on the added worker effect, this idea of women entering the labor market in bad times. And with this, we are contributing to the literature on scarring effects, the analysis of the long-term consequences of entering the labor market during a recession, which is very extensive for developed countries, but very limited for developing countries. And we add another result. What we are going to say is that there is a group of workers who may end up benefiting from entering the labor market during a crisis or in a bad time. And this is the group of young women. So more exactly what we are going to do is to analyze the effect of facing high unemployment rates at the moment of presumed labor market entry. On workers' labor outcomes, the role of women within the household and perceptions about gender roles in society. All of these measure 10 years after entry. So our outcome variables are measured during adulthood. We have a group, a set of labor market outcomes that comes from a cross-section of national household service from 15 Latin American countries over the period 2001-17. Then we have two measures capturing the role of women within the household. These measures also come from household service. Then we have a set of variables capturing perceptions about gender roles in society. They come from work value service and Latino and Aumetro. The service what they have is a set of questions about agreement with traditional gender norms or stereotypes. So a typical statement in this kind of service is when a mother works for pay, the children will suffer. The person will say if he's agreeing or not with the statement. And our explanatory variable is the national unemployment rate at the moment of labor market entry. I will define labor market entry in a moment. This information comes from work development indicators and national household service and covered the period 1992-2007. That was a very unstable period for the region. With a lot of variation in the unemployment rate across countries and over time. So how do we define a labor market entry? Here the common practice in papers from developed countries is to use the actual date of labor market entry or actual date of graduation from college. What we are going to do instead is to use age to create a proxy of labor market entry. And we are going to use the calendar year in which the adult persons we see in our data were between 18 and 20 years old. We use 18 years because this is the theoretical age for completing secondary education which is compulsory in most of the countries of the region. And where we see an important jump in the rate of labor force participation. Here in this figure we have the average for the 15 countries in our analysis. And we have that the labor force participation at ages 15 to 17 is only 30%. But for those 18 to 20 is almost 60%. So we think that using this age range we are capturing the school to work transition. Now what about adulthood? What we are going to do is to construct a year of birth cohorts using the cross section of a household service. We focus on cohorts that were born between 1974 and 1987. That gives us 14 cohorts in each of the 15 countries. And we are going to observe these cohorts over the period 2001 to 17 when these persons were between 27 to 30 years old. And the goal here is to analyze whether the outcomes we observe at ages 27 to 30 depend on the labor market conditions that these people face at ages 18 to 20. So take for instance the first cohort in this table. These are persons that were born in 1974. We observe their outcomes between 2001 to 2004 when they were between 27 and 30 years old. And these persons presumably entered the labor market when they were between 18 and 20 years old. And that happened between 1992 and 1994. So what we are going to do is to calculate the average unemployment rate between these two years. And this is going to be the shock affecting the outcomes of adult persons in our data. So basically what we are going to do is to use the variation in the unemployment rate across countries and across cohorts to identify the effect of entering the labor market facing adverse conditions that is facing a higher unemployment rate at ages 18 to 20 on outcomes during adulthood between 27 and 30 years old. And what we do is to follow the literature and instead of working with individual level observations we are going to use as our unit of analysis cells or groups defined by cohort, year, gender and countries. So all our variables are defined as averages within each cell. And we combine all the information for the 15 countries. And we propose following the literature a model with fixed effects by country and years, estimated by all this. We estimate it separately for men and women pulling all the information for the 15 countries. Are you seeing the countries? In one of the exercises we are going, they can go to college and in that exercise we are going to change the age. Instead of being 18, 20 it's going to be 22, 24. Yes, yes, yes. And that would be the case if we were using the actual date of entry to the labor market or actual date of graduation. But we are using age. So the composition of the cohort is likely to be exogenous because it only depends on year of birth. So going to the results. Here we have the first set. What we can see easily is that the pattern of results is very different for men and for women. For men, the story is very much in line with results for developed countries, men from the unlucky cohort, those facing higher unemployment rates at ages 18, 20 and being penalized some years later. For labor force participation, we have a negative point estimate for men, a positive for women. They are close to be significant but they are not. For unemployment the same, they are not significant. For employment, we do have significant results. We have a reduction for men and increase for women. And this gender type of gender difference didn't appear before in papers from developed countries. And here in the other labor market outcomes, we have no effect on ours but we do see an increase both in early wages and monthly earnings for men and for women. For men, we were expecting this increase because of the composition effect. We expect low productivity workers to be the first one to be discouraged to enter the labor market or to lose their jobs. For women, we were expecting the opposite. What we think might be happening is that this increase could be connected with changes in social perceptions about the role of women at home and in society that allowed women to advance their labor market careers. Now, how do we explain the increasing employment for women? We think a possible explanation is the other worker effect. So the women enter in the labor market in bad times. But the thing is that our result in increasing employment appears 10 years after entry. So what we think is going on here is that these women enter the labor market at the moment of the job and then they remain in the market. So what we do to test that is to run again the model but now using a outcome variable, the labor force participation at ages before adulthood. So we have here the estimation for different age groups. And what we have as result is for men, we see always a negative point estimate that sometimes is significant. For women, we have the opposite. The point estimate is always positive and sometimes it is significant. So we think this is evidence consistent with the other worker effect. So the idea would be that women enter the labor market at the moment of the shock or immediately after the shock but they remain in the market and we observe this effect on employment 10 years later. So what we need to explain now is why the persistence of the effect, why they remain in the market. And what we think might be happening is that there might be some changes in behaviors or perceptions once in the labor market. So the idea here is that once in the market, for instance, women may change the value, they assign to having a job and being economically independent or there could be a change more generally at the level of the society in the attitudes to our working women. So what we do to test for that, what we do here first is to use as outcome variables to measures related to the role of women within the household. And what we see here is that for higher levels of unemployment at ages 18-20 leads to women to control a higher share of the household labor income and women to be more likely to be the head of the household. The first effect is impact mechanical. We saw before an increase in employment and in earnings but this is a measure that is usually associated with increasing bargaining power. So what we think is that these two results are showing us that women from unlucky cohorts increase their bargaining power within the household 10 years after the labor market entry in comparison to the other cohorts. And in the last set of results what we do is to test this idea of changing perceptions about gender roles but at the level of the society. So what we have here are outcome variables that are binary indicator of strong disagreement with statements reflecting traditional gender roles or stereotypes. And what we have here is that facing higher unemployment at ages 18-20 lead to increases in the disagreement with these statements. All of the point estimates are positive. We have two significant results for the first two measures that tells us that the statement is that there is a problem if women have more income than the husband. And the second is when a mother works for pay the children will suffer. So what we are having here we are finding changes in society's attitude towards this traditional gender role. What we have is that for people from the unlucky cohorts there is an increase in the percentage of individuals that strongly disagree with these gender stereotypes. And at the end we run some other models and some robustness. First one of the possible channels behind the positive result for women are improvements in education. So what we do is to use as an outcome variable years of education. And what we find is that there is an increase for both men and women. But what we think is that this cannot be the channel explaining the results for women because if that was the case we should be having similar results for men and we are seeing the opposite. So what we do is to estimate again all the models but now controlling for years of education and we find a similar conclusion. So the main results remain even after controlling for education. Then a second thing we do is to instead of using the standardized measure of unemployment to use the unemployment rate without standardizing and again we obtain the same set of the direction of the results is the same. Then we do what Raquel was asking we change the presumed age of labor market entry according to the educational attainment of adult persons. So for persons who are 27 to 30 years old with secondary education we use the shock at ages 18 to 20. But for workers who have college education we use the shock at ages 22 to 24. And we still see the same set of results in terms of direction and patterns of effects. And in the last exercise we do is that we change the age of adulthood. Instead of looking at ages 27 to 30 we use a broader definition of adulthood and now we look from age 25 to 35. So we have a bigger sample there but it is, if you remember the figure I showed before. Here, when we use ages 27 to 30 we are working with what we say is a balance panel when we work from 25 to 35 the panel is now unbalanced. So that was the last slide. So to wrap up what we have analyzed here are gender differences in the effect of facing adverse conditions at the moment of labor market entry in 15 Latin American countries. We found very different results for men than for women. The story for men is very much in line with evidence from developed countries. For women we found the opposite. We saw increases in employment and in earnings. We think this is related with the other worker effect so women enter the labor market at the moment of the shock and then remain in the market and the persistence we think could be related. We have suggestive evidence that it might be related with changes in perceptions about gender roles both within household and at the level of the society as a whole. Thank you.