 I'm going to talk in general so we kind of know and this session is about affirmative action and we know affirmative action is important and it's been used in a lot of countries. So what I'm going to talk about is two things. The first one is I'm going to talk about two papers I have on discrimination and their effect of affirmative action in India, both using lab in the field experiments. And then I'm kind of quite excited about the second aspect, which is what we're going to do is we're going to test a model and then we're going to describe and try and hypothesize about why affirmative action works in some places and doesn't work in other places and then we can try and understand what we can do to kind of improve the effectiveness of affirmative action. So the first one is what we kind of mentioned a little bit before so this is already published in European Economic Review so I'm not going to spend too much time on it but I think it's kind of nice to just talk a little bit about this but it was also funded by UNU Wider. So in this paper what we're trying to do is to try and understand do men and women respond differently to women as leaders and then what we're trying to understand is is there an effect of gender quotas on the how people behave towards female leaders right and so what we do is we run a lab in the field experiment that is specifically designed to answer these questions and I'm kind of interested in lab in the field experiments because I believe they have a lot of control and we can really kind of get a behavior and understand behavior in a very controlled setting. So we do this in India and we know from the last week in the Indian context we have quotas and we have gender quotas and also quotas for caste in this case we're looking specifically at gender quotas and so what we do is we examine the behavior response to women as leaders as distinct from the impact of leaders and then we examine the channels by which quotas affect behavior in this context and then we also observe the behavior of men and women towards leaders. So to do this what we do is we run a one-shot public goods game and this game just measures cooperation and basically people that don't know I'm not going to talk about in any detail but people when they play this game this is in villages in Bihar in northern part of India they can contribute towards a public good or they can contribute towards a private good right and in each group there's two women and two men and then for each group there's a leader who's either a male or a female right so then people within that group can contribute towards a public good that has a leader that is a male or a female right and people know this so one thing that's important is that the leader proposes a non-binding contribution about what they suggest people should do towards the public good and then the leader's proposal communicated to the group members and then all group members including the leader contribute towards that public good and then payoffs are calculated so we have two treatments one where the gender is revealed the gender of the leaders revealed and one where the gender is not revealed and so this is run in the context of the affirmative action policy in India where we have 50% of all village chief positions and technically all positions are reserved for females so what we can do is we can look at how people behave towards a female leader in our game in areas where there's a quota in areas where there's no quota so what we find is that this is the general result we find that men contribute less towards female led groups it's a negative so that means they country less towards female led groups and then we also have this as female group members and they can come they contribute more towards female led groups but male group members contribute less towards female led groups and is this the same in female villages where there's an So in villages where there's a female chief. Well, so what we find here is this male citizens, this is where there's a male chief and this is where there's a female chief. So male citizens only contribute less towards female leaders when there's a female chief, right? And then female citizens contribute positive amounts in both. So this is consistent with what we found, what the last speaker found. And this is basically evidence of sabotage but evidence of sabotage when there's a female chief. So experience with the female chief. So that suggests that affirmative action was not very effective in that context but I'm going to talk more about this and come off hopefully in a positive note. So the second study I'm going to briefly talk about is a study with Sonja Brilotta, Irma Klotz and Lakshmi Aya and this is more of a recent study. I think we now have an Isaiah working paper out and this is really similar to what we've done in the last paper but I'll talk about the differences. So in this paper what we're interested in is we're interested in not in cooperation which is what we're interested in the last paper but this paper we're interested in coordination and in particularly coordination failure. So this kind of goes back to what Kashik Basu was talking about earlier but in this case we're interested in Hindus and Muslims. So what we want to know is that we want to know if leaders and if a Hindu leader or Muslim leader can improve coordination in a village and if it can bring us out of an equilibrium where we have coordination failure, right? And so what we do is we test that but then what we try and assess is we're trying to understand what is the effect of that leader when there's an affirmative action policy for one of those leaders and also what is the effect when there's intergroup contact. So this is where we test the contact hypothesis. So what we try and do in this paper is we test affirmative action relative to the contact hypothesis and we also do this in the context of areas that have experienced conflict, religious conflict and areas that have not experienced religious conflict. So we're trying to test both of these policies generally but also these policies when there's a history of conflict, okay? So we do this in UP in India with just over a thousand individuals. Again, this is a lab in the field experiment. We do this in UP because there's a lot of history of religious conflict in this state. We run four tasks. I'm really gonna briefly go through the important things again and what we do is we have three treatments, one for each policy, we have a control, then we have a treatment where there's an AA policy and there's a treatment where there's a contact policy. And we do this in what is called a weakest link game which I'll explain briefly and everything's registered with the AA registry. So we're not just looking at conflict because it was significant in the end. We prior registered conflict as our interest. So we've specified our sample to be interested in conflict prior. Okay, so let me just really briefly go through the task. So those that are not familiar, this is a weakest link game. It's based on the AR paper of Jody Brantz in 2006 and it's six periods long. It's been posed of two stages. In each group of four, we have two Hindus and two Muslims. So this is done in towns in UP. This is not villages because we need slightly more educated people and also because conflict often occurs in towns. We have two parts. We have periods one to four and periods five to six. So in periods one to four, this is how we set up the game. This is a game. Subjects are employed in a firm that has four people. They must decide how many hours to work. This is decided effort, not actual effort. So this is exactly how the game was set up in the AR paper. We haven't changed it at any point since at this point. Payoffs depend on own effort and the minimum effort of others and informed of the minimum effort after each period and coordination in this game is set up so that it's very, very difficult. There's no leaders so far in this game but what happens is that coordination is very difficult because your payoff is dependent on the minimum effort of the group as well as your effort. So let's have a look at the payoff table here. So this is the minimum hours of the other three employees in your group or technically all the employees in your group and this is your hours. As you can see, if you increase your hours but the minimum hours do not increase then your payoff will decrease, right? So this is why it's a coordination failure or this is why we have a coordination problem. This is why this is a coordination game. So what happens is effort is costly. Subjects payoff is an increasing function of the minimum effort chosen by the other group members. So we can show you theoretically that you should only increase your effort if you believe the minimum effort will increase and it's pretty obvious that that should be the case. So in other words, periods one to four this is how we set up the game. We induce coordination failure. In nearly all cases, we get a coordination of zero as the minimum effort, right? And then we wanna introduce a leader in periods five to six. We wanna introduce a Hindu leader and a Muslim leader and we wanna see do these leaders offset coordination failure. So we have a group leader from these four that is randomly chosen. Again, similar to my other paper, each period out of period five and six, the group leader must suggest the number of hours to work. Citizens are informed of this suggestion and they're also informed of the identity of their leader. And then all subjects must decide how many hours to work. Again, this is decided effort, it's not real effort. So then what we do, that's our control. We have no policies. Then we have a treatment where we have affirmative action and this is simply, it's more complicated than this but simply when we introduce a leader we also tell them that if they have a Muslim leader, we say that your leader is in the position because of an affirmative action policy. This is a reserved leader position. If it's a Hindu leader, we say it's unreserved, okay? Then in our contact treatment, we have people play Hindus and Muslims played together in some interactive task that's non-cooperative before the game. And they do this about 15 minutes and they're encouraged to coordinate. So this is just what happens in the affirmative act. This is the minimum effort and this is Hindu leaders and this is Muslim leaders when we have affirmative action. So what we find, if we compare this to control, the control has discrimination where there is differences but this is much wider. So in other words, when we introduce affirmative action what happens is Hindu group members behave really badly with Muslim leaders, right? So we have sabotage again, yeah? And this is bigger than the control. And this, so basically here, Muslim leaders don't really have any impact on coordination but Hindu leaders do because Muslims don't have backlash against Hindus in this case but Hindus have a lot of backlash against Muslims, okay? So then what happens in the contact? Well, this is the opposite. Contact basically the differences between the two groups go away and we have much higher minimum effort, okay? So that's what we kind of found so far. There's much more results and you can kind of read the paper. This is kind of a summary and what the key thing I'm trying to kind of represent here. And what I'm really interested in and what I wanna kind of bring to your attention is this paper which is now conditionally accepted at Management Science. And so in this paper, what we're trying to do is we're trying to understand why affirmative action works and why it doesn't work, okay? And we have a model and then we have a large sample from the US and then also an experiment where we test this theory. So this is what we try and do. So to kind of explain this not using math, what we argue is that gender quote is a controversial, right? Some people argue that they are unfair. Why? Because they say the best person's not getting the job, right? You've probably heard this before and I think this is one of the arguments in India. The best person's not getting the job or the best person's not getting the position so it's unfair, it's reverse discrimination, you've heard all these things before, right? Other people would say they are necessary because females or minorities have to go the extra mile to get the same recognition. So there's some form of disadvantage and we need affirmative action to bring them to an equal playing field, right? So we have these two arguments. There's more, it's more complicated than this but I think these are kind of the two arguments. And what we argue is that these revolve around the best person for the job. So what is we term meritocracy, right? So therefore we propose that whether quotas is meritocratic depends on the perceptions of the environment. This could be the environment in the village, this could be the labor market environment, okay? So the simplest way to represent this is this diagram. So we believe that these are two important factors that lead to the differences in the effectiveness of affirmative action. This is what we call the disadvantage axis and this is where there's a bias against women and women have to work harder. I'm talking about women here where you can talk about caste, you can talk about minorities, religion, it's the same, yeah? So along this axis, this is where you have disadvantage. So the further along this axis you go, the more disadvantages there are against women. So women have to work harder, yeah? So this is based on beliefs. I'm talking about beliefs here. Similar to what Kashi was talking about, we're interested in some sense about beliefs because that's what's driving a lot of this behavior. Along this axis is what we call gender skill gap belief, okay? So this is where people believe that there's a skill difference between males and females, or minorities and other majorities, right? So along this axis, people believe that females are less skilled, this is perception, again, than males, right? So the further along you get, the belief changes so that people believe that females are less skilled than males. At this axis here, people believe there's no disadvantage and no skill difference, okay? You could be at some point here, you could be at some point here, you could be at some point here. Let me just kind of go to kind of an interesting graph. So this is a large survey in the US where we surveyed people's occupations. So this is arguably a representative survey that we hired a survey company to do and we asked them, in your occupation, do you believe there's a skill gap? Do you believe there's a skill difference between males and females? And we also said, in your occupation, do you believe there's disadvantage for males and females in leadership positions? And these are just different occupations. And you can see that within these occupations there's differences. This is the military and security occupation. People who work here believe there's a skill difference and not much disadvantage. People who work in transport believe there's a skill difference but lots of disadvantage. People who work in construction believe there's disadvantage and a bit of skill difference, right? What I'm trying to establish here is that these are occupations but this could also be societies. You could plot societies on here, you could plot, say, Finland and maybe they will be at some point here. You could plot India and perceptions could be at some point here, right? But what we argue is that this is important and this can kind of explain, these beliefs can explain the effectiveness of quotas. So then what we do is we run a large survey in the U.S. of just over a thousand U.S. residents, representative and we asked people this question. We asked them based on these four different characteristics. Based on, we're looking at the axes here because it's easier to look at the borders. We're interested in this case, this case and this case, right? We can also be interested in this case but it's harder to explain so we're not gonna talk about that now. So we ask people, standard, what do you think about gender quotas in the U.S.? Do you think they're appropriate or not appropriate? So we have kind of a little bit of a distribution where we have about 45% of people say it's appropriate to have gender quotas for leadership positions and about 40% of people say it's inappropriate to have quotas for leadership positions. But how about if we have these three different environments? What happens? Let's look at the first environment. How about we create a situation where women are on average less qualified for leadership positions and there's no bias, there's no disadvantage? Do you think quotas are appropriate? Most people say no, they're not appropriate. 60, 70% of people say they're not appropriate, right? How about the situation where they are equally qualified for leadership positions and there's no bias? This is the origin, yeah? No bias and no skill difference. Do you think quotas are appropriate? Most people say no, quotas are not appropriate, right? In the U.S. How about the situation where there is no skill difference but there is disadvantage? Women have to work harder. Are quotas appropriate here? Well, nearly 80% of people say, just less than 80% of people say, quotas are appropriate in that case, right? So in other words, what we do, what we have when we kind of manipulate these beliefs, we kind of have more consistency. But that tells us nothing about economic impact. We don't have any causality here. So we run three experiments in Australia where we wanna test each of these. We wanna have a situation where there's a quota here and a no quota here. A situation where there's a quota here, no quota here. And a situation where there's a quota here, no quota here. And we do this, and how we do this is we manipulate beliefs in each of these situations. Where we say in this experiment, we tell people women are disadvantaged and we create a situation where they have to work harder. In this situation, we say there's no disadvantage and no skill gap. In this situation, we create a skill difference, right? And we run what is called a gift exchange game which is a principal agent game where we have managers and workers, some of the manager positions are reserved, some of the manager positions are non-reserved. We have these two treatments, right? And we have a sample of about 600 people in Australia. And so what we find is basically consistent with the survey. We find in our experiment, in our gift exchange principal agent game where we have female quotas. In this situation, female quotas have much worse welfare outcomes in terms of the game, right? Relative to a situation where there's no quota, right? So it's just meritocracy. Then we have a situation here where we have our second experiment. What happens? Again, we find that in the situation of a quota, welfare in terms of our game, so we can talk about this in terms of effort and also manager wage are much lower in the quota situation relative to the no quota situation, okay? Again, consistent with our survey. But in the situation here where there's disadvantage, we find that quotas have much higher welfare outcomes than no quotas, okay? And this is based on what we talk about in our model, which is meritocracy. This is all based on beliefs about the fairness about the appropriate person being in that position. In the situation here, it is fair to have a quota in the sense that the best person is not getting the job because they're disadvantaged. So therefore, quotas is more appropriate or it's more meritocratic in that situation. In these two situations, quotas are not necessarily meritocratic because there's no skill differences or there's no disadvantage, right? And therefore, quotas can arguably mean that the best person's not getting the job, right? And so as I've mentioned, quotas have better outcomes relative to no quotas in this disadvantage experiment. So just to conclude, what we find is that in our two situations in India, quotas have certain backlash, generally always driven by the person who's the group who's not part of the quota setting. But what we find in other settings and as long as you believe this belief is important, you should be able to transfer to other settings. But what we believe is that beliefs about meritocracy, so in other words, if he will believe that that person is disadvantaged or that group is disadvantaged and that deserve the quota because they have to work harder, then quotas can be effective. But in situations where people do not believe quotas useful because they do not believe that there's disadvantage or that they believe they deserve to be in that position, then quotas maybe have some negative effect because the other group is upset and will create some backlash. And I think that's it.