 Hello, good afternoon, thank you very much for the invitation. So we've seen a really like context specific, country specific studies, but here we will zoom out and focus on the whole Africa. So this is a joint work with Luisito Bertinelli and Rana Comapay from the University of Luxembourg. And I need to acknowledge the support from the World Bank and UNHCR under the programme of Preventing Conflict and Promoting Social Cohesion in Fourth Displacement Context. So the question we are asking is simply to know how do refugees affect conflict through one particular channel which is changing ethnic diversity. And our contribution is twofold. First of all, there is a large literature on the relationship between ethnic diversity and economic development in general, but in particular related to conflict. And certainly the seminal papers, both from a theoretical point of view and empirical point of view are the papers from Esteban and Ray encoders. And basically what they say is that you need to distinguish ethnic fractionalisation from ethnic polarisation. So to make it short, ethnic fractionalisation is the probability that two individuals would belong to two different groups. So the index will be maximised when you have a lot of groups to the extreme theoretically when you have one group per person. Polarisation is different. It captures antagonism between groups. And it will be maximised when you have two distant groups of similar size. So there are really two different measures. And what Esteban and Ray are saying is that basically polarisation should increase conflict while fractionalisation should matter less because it increases coordination costs. So far the evidence has been mainly using cross-country analysis. We see two limits in that. One is that this type of index, they move very slowly over time. So you lack a bit of time variation. And also it's more difficult to control for unobserved characteristics. So in this paper what we are going to do is exploiting time-varying diversity indices at the sub-national level induced by refugees inflows. So the second contribution is that we contribute to the recent literature that has assessed the impact of forced migration on hosting societies. So this literature has looked at a lot of incomes and we can discuss that. But regarding conflict, Christopher mentioned a bit this literature. My reading of the most recent evidence by Yang Yang Zhu and Shaver is basically that there is no direct impact of refugees on conflict. And there is a more recent paper showing a very short-lived impact. What our paper is showing is that it does not contradict at all these findings. It's just saying that in the particular case where it changes ethnic composition, refugees may actually exacerbate conflicts when it increases polarization and reduce conflict when it increases fractionalisation. So what we are doing, it's easier to explain you the construction of the dataset. It's a bit complex, but let me give you the intuition. So we are using available data from the Afrobarometer for 23 countries. So we have all the clusters, if you want the villages of the Afrobarometer, then we create buffer around each cluster. And we count to some extent either the conflict incidents or the conflict intensity. And basically from the Afrobarometer, we can compute the diversity indices of fractionalisation and polarization. Okay, so so far there is no refugees. Then basically we introduce a dataset on refugee camps in Africa from 2005 to 2016. So we have the exact location of the refugee camps, but what is interesting for this paper, we have also the composition of these camps. Age composition, country of origin, and combined with the ethnic power relation, we can approximate for the ethnic composition within camp. So now we are going to recompute the diversity indices, fractionalisation index, polarization index, including the changes in ethnic composition induced by annual flows of refugees at the camp level. Okay, I grab that. Okay, so from there, you should more or less see what we want to do. So we follow the game theory contest model between groups by Esteban and Ray, right? And we are going to try to explain whether the incidents or the intensity of conflict, right, in one particular cluster, J, in your T, is influenced by the refugee corrected ethnicity indices, so fractionalisation index, polarization index, controlling for the presence of refugees. We are going to control for unobserved characteristics using cluster fixed effect and your fixed effect and controlling for climatic anomalies, right? So this is a simple or less estimation. Of course, one concern we may have, right, is that refugees self-select location according to ethnic characteristics. So I'm going to argue that that's really the most important is the selection with respect to ethnic composition. And a priori, it's quite difficult to know actually how it's going to affect ethnic polarization or ethnic fractionalisation. There is no clear conjecture on that. So what we are going to do is we are going to apply a gravity model estimated with PPML, where we are going to predict what the number of people, right, from one particular ethnic groups moving from one location to another in one particular moment in time, and we will recompute the diversity indices and use that as an instrumental variable, right? Okay. So the main results, as you can see in the first two columns, we are using the standard fractionalisation and polarization index, right? There is no significant result which suggests, you know, that there is very little variation there. In column three and four, we introduced this corrected diversity indices, right? And as expected, we do find a positive impact from polarization, right? And a negative impact from fractionalisation, controlling or not for the presence of refugees, controlling or not for climatic anomalies, right? So just to give you a sense of the magnitude of the results, right? A one-star deviation increase in polarization will increase the risk of conflict by 5 percentage points, which is about 10% at the mean, right? And interestingly, it's quite similar to what you have found in the literature regarding the size of the effect from natural resources, from prices, from external shocks like climatic shocks, right? So it's quite sizeable as an effect. And we do not find any significant correlation, I insist correlation, for the presence of refugees, okay? So we do quite a lot of robustness checks. So first of all, we look at different type of conflict. What our result seems to suggest is that it's mainly driven by violent conflict, right? But it's also valid for intensity, but we have to say it's more robust for polarization rather than fractionalisation. We do a bunch of robustness checks. You know, there is a lot of work in matching ethnic groups. So we do some alternative way to match them. We also change the buffer around each cluster from 40 to 120 kilometres. There are some new things that are not included in the paper, sorry, Arzu. So for example, we control for conflicts below over using the distance to the border times the time fixed effect. Something which is, I think, interesting is that so far we did not, we assume that all groups are different, right? So now what we have done now, and we are changing the old paper, is we introduce intergroup distances, right? So using linguistic, linguistic distance, right, between groups. And we can also show that the results are pretty similar if we aggregate the data at the regional level, rather than working at the cluster level. So I'm not going to enter too much into details, but for the instrumental variables, we do find similar results, right, when you instrument. The coefficient for polarization is actually higher, suggesting that the ALS estimate is underestimating the effect. And so I'm running to my conclusion. Oh, okay, so I'm even faster than... Okay, but so I think what I want to insist on is one thing, right? Our results should not be misunderstood in the sense that refugees per se do not exacerbate conflict, right? We do not contradict what has been found, for example, by Yang, Yang, Zhu and Shaver. But we show that in the particular case where ethnic polarization increases, the risk of conflict exacerbates, and the opposite for fractionalization. I did not take the time to show you that, but we also confirm this result using individual data from the Afrobarometer, right, using physical assault or interpersonal violence as a dependent variable. And so we are still working on that, but it's quite interesting to see that it's not related to other outcomes like ethnic attachment, generalized trust, trust in neighbors, or institutional trust, which tend to suggest that we are more in the Esteban and Ray framework than in other type of framework. And maybe from a policy point of view, it's quite interesting to observe that this heterogeneous... This is partly true when you desegregate between unemployed and employed people. So unemployed people are partly likely to participate, to report physical assault, right, when you use individual data. So policy conclusion, to be honest, that's not the paper with the strongest policy conclusion, right, compared to others. But I think there are quite a lot of specific intervention that, for example, the World Bank tries to promote to and social cohesion between the host and the refugees, which is welcome, but certainly there might be a need to target this type of intervention to highly polarize hosting areas. And maybe there is a need to map systematically or collect data on ethnic diversity among the refugees and the host to follow, actually to be much more context specific than what I give you today. Thank you very much.