 Thank you, thank you very much for allowing me to be here. This paper is called Ethnic Conflict, The Role of Ethnic Representation, and it joined work with Sonia Balotra and Laxmiaya. I still work in progress. So, I guess you remember several incidents that took place in India in 2020, in Delhi, in Bangalore. They were calling them communal incidents, even if what they refer when they talk about that, they refer to violence between religious groups. And in those incidents, at least 73 persons lost their lives and many others were injured. Religious conflict is not something that only happens in India. You all know that it's an important component of many within-country conflicts around the world. And in India, however, the majority of these conflicts are between Hindus and Muslims. Hindus are around 79% of the population, and Hindus form around 14%. So, there have been some incidents of special relevance with a large number of injuries and deaths, like what happened in Gujarat in 2002, where some Muslims attacked a train where some Hindu pilgrims were, and also in Atapredesh in 1992, where some Hindus demolished a mosque in Atotika. So, it goes in both ways, and there have been some incidents of special significance. What we want to understand in our research is whether ethnic political representation affects ethnic conflict. In particular, whether the presence of Muslims in state legislators in India change the probability of Hindu-Muslim violence. And then in order to analyze this question, we create two datasets. One is on Hindu-Muslim violence. We extend the Barney Wilkinson dataset from 1995 to 2010, and the other one is on the religious identity of state legislators that we have already used in some other papers. Now, since the presence of Muslim legislators is endogenous, or we cannot randomize Muslim legislators, we take advantage of the fact that some of them win in close elections against non-Muslims, and we use a regression discontinuity design. Today, I'm going to use an approximation to a regression discontinuity design using two-stage list squares with the data that we have. So, India gives religious freedom and allows Muslims to propagate their religion. There are minorities still. There are 14% of the population, even if India holds the largest Muslim population in the world, and they're a bit different from Hindus in the sense that they're more likely to live in urban areas. They're poorer, but they display better health outcomes. So, why do we think that Muslim politicians could have an effect of inter-religious conflict while there are several mechanisms that could be taking place? Previous literature suggests that Muslims have been the ones that have suffered more from conflict, so then this means that maybe security or lack of violence is a public good that Muslims like to have, so politicians may be more likely to invest in that public good and to provide security. How could they do that? Well, they can mobilize the state machinery, for example, use the police, control tensions before the violence erupts, or maybe use community-based institutions to resolve these disputes. And alternative, however, maybe the presence of Muslims in public office generates backlash, and then that increases conflict. That would be consistent as well with previous literature that suggests that when Muslims do well economically, violence increases. So, when they have more political power, maybe violence increases as well. However, there is very little work on what is the effect of political representation on minority groups' violence in the world and in India. So, we use data first on politicians. India has a parliamentary system of government. Politicians are elected in constituencies, and there are no Muslim parties, very few, and even though Muslims tend to belong to the Congress party or largest parties. We created this new database on the religion of politicians, basically classifying politicians by name. We used a program called Nampatchian to do that, and we also had a team in India classifying the name. So, whenever there was a disagreement, we resolved it. Ourself, on an average, 7.7% of state legislators were Muslims over the time period, and the consideration one-third of districts had at least one Muslim legislator. If you look at how Muslim political representation has evolved over time, there have been, it's been more or less, more than even if there was a bit of a decrease in the 90s, but it's smaller in percentage, that the percentage is a population. The data on religious conflict is obtained by updating the Pasne Wilkinson database. This is data that comes from the Mumbai edition of the Times of India. We basically followed their methodology and updated it from 1995 to 2010. So it's basically reading newspaper articles and trying to quote what is a religious incident between Hindus and Muslims or Muslims and another group. However, we don't have incidents between Hindus and Sikhs or Hindus and Christians. So that would be two examples of newspaper articles that we used to classify them. In most cases, the article provides a location of the incident, say the district or the town, and that's what we're working on now. So we are basing our analysis on more than 1,000 incidents that took place between 1980 and 210. For a large majority of them, we've been able to find which district the incident took place, and we're working on the town data. So far, however, there are 186 incidents that we cannot classify because they took place on a large city, and we would not know in which electoral constituency that incident took place. So today I'm going to present the district data, but we are working on the town data, and we will have it very soon. So if we look at the trends in ethnic violence, they have decreased over time. From the 90s, it has gone down a little bit, apart from some larger spikes, and it has happened across all states. So if you look at this graph, it has decreased in all states, but it's prevalent almost everywhere, even if it tends to happen in states such as Gujarat, Maistrastra, and Uttar Pradesh. Now our question is whether political representation affects religious violence, or how do we identify this? Given that Muslims and non-Muslims may be elected in areas that are not comparable, or maybe Muslims tend to be elected where the Muslim population is larger. For example, we cannot simply run an oil as regression looking at the percentage of Muslims elected in conflict. So we take advantage of the fact that some Muslims won't enclose elections by very few votes against non-Muslims. So then in a constituency where a Muslim won by very few votes against a non-Muslim, we can assume that preferences and the setting will be very similar as in another constituency where a non-Muslim won by very few votes against a Muslim. So they both will have been elected in similar circumstances, and this difference in votes could be interpreted as random or quasi-random, as we would like to call it. So then this regression discontinuity would allow us to estimate the causal effect of a Muslim political victory on violence. Now in the current setting, we implement it as a two-stage least squares regression because we don't have information on the particular constituency where the Muslim was elected and the conflict took place. So we have districts where there are around 10 constituencies and we know that the conflict has taken place in that constituency, in that district, but not in which particular constituency. This is the next step what we're currently working on now. So then what we do is we use regression discontinuity as a two-stage least squares. So we basically instrument the fraction of seats in the district won by Muslims with a fraction of seats in the district won by Muslims in close elections against non-Muslims. So once you control for the existing of these close elections for the fraction of districts that had these close elections, the fraction of Muslims that won in close elections would only be related to violence by affecting the fraction of Muslims elected. So then the exclusion restriction would be satisfied and we would be capturing the causal effect. We obviously conducted one million robustness checks first to make sure that we could use regression discontinuity and that the instrument works and so on. And I'm not going to show you this and I'm not going to show you all of the results but some of the main ones that are robust to many specifications. So here you have the dependent variable what we're interested in is number of incidents in the first column. And we see that the fraction of Muslims elected instrumented with a fraction of Muslims in close elections decrease the number of incidents. If instead of looking at the number of incidents we look at the probability that there has been any incident, we also observe a significant decrease. And then when we look at any injuries or when we look at the number of deaths we find negative coefficients even if they are not statistically significant. So Muslim political representation seems to decrease Muslim violence and this seems to be robust. We've conducted many robustness checks as well. Now if Muslim political representation reduces the number of incidents and the probability that an incident takes place at the district level and it's robust we need to understand the mechanisms at play. So our next step is first of all look at the town data. The town data will allow us to identify in which particular constituency the conflict took place. So then we would directly relate a politician with a conflict incident but also we want to know what variables such as police officers, police officers that are Muslim. What do these politicians do in order to stop the conflict because it seems like they stop it and that would lead to questions such as should there be quotas for Muslims if actually improving the political representation of the minority reduces violence and how should this be done? But that's a future research and thank you. Thank you very much.