 In terms of what we expect to find, there's a bit of literature mostly on developed country context on this. So in terms of whether we would expect refugees with disadvantage to play a role in support of redistribution, there's a lot of literature on developed countries suggesting that primarily it's the self-interest of individual voters that plays out in their decisions to support or not redistribution. And so given that the budgets are fixed, people often opt for support to things that they would benefit from. On the other hand, there is some literature to suggest that these considerations of fairness and altruism play a role in decisions to support or distribution and also perceptions of fairness of social competition. For example, there are papers that show that reason why European countries have a larger welfare state compared to America is because Europeans are more likely to think that luck in life determines more compared to Americans who are less likely to think so. So we don't, you know, probably have a certain expectation on how the perception of disadvantage would, whether they would play out in people's decisions to support a distribution or not. And in terms of whether refugees feel part of the society, how integrated refugees are and whether that matters for people's choices to support government assistance or in favor of refugees, we know that the group loyalty and social identification are important in preferences around redistribution. So people are more likely to support distribution in favor of those to who they relate to, could it be as defined by ethnicity or things like that. But then we also know that these things are different depending on political economy and economic context. And also we know that integration has multiple dimensions. And so depending on whether the integration is perceived with respect to cultural identity or whether it's respect in terms of economic integration in terms of whether the people work or not or they're unemployed, these dimensions of integration are important. So we would like to think that we contribute to the literature in some ways in the sense that even though there are studies on redistribution preferences in general, they're mostly focused on developed countries and there's a lack of developing country evidence. So we provide some of that. And also we are looking at specific dimension of redistribution, specifically looking at government assistance to refugees. So refugee targeted assistance. We're also interested in perceptions. So perceptions are of course objective and it's hard to work with perception based data in statistical context. However, we also know and we also know that perceptions may not necessarily coincide with the reality. So if you are looking at refugee disadvantage, it may not be that individual perceptions of how refugees are disadvantaged may coincide with how indeed refugees are disadvantaged relative to the mainstream society. However, perceptions are important in people's, a lot of important decisions that people make and they may indeed be more important than objective data. So we like the idea of contributing to the research that deals with perceptions. Quite a bit of literature on studies on attitudes to immigrants and our measure of whether people want to support government assistance to refugees or not can be thought of as a proxy of attitudes to immigrants. However, this literature is mostly concerned with labor migration and there's less evidence on refugees and there seem to be the case that attitudes to refugees are different to attitudes to migrants. In some cases, interestingly, they can be positive, more positive than those two migrants. And again, there is a lack of evidence on developing countries. And finally, we are not going to talk a little bit now about the specific context where I say interested in but refugees are often the result of conflicts and in this case, we will be looking at Caucasus and there's a relatively limited literature on the consequences of conflicts, particularly around the resulting displaced populations. So we are looking at three countries in South Caucasus, Armenia, Azerbaijan and Georgia. These are formerly Soviet Union countries. So there have been a number of ethnic conflicts and wars that erupted following the Soviet Union collapse in Nagorno-Karabakh between Armenia and Azerbaijan, South Ossetia and so forth. And as of the beginning of 1990s, the estimated number of refugees and IDPs in this region was around one and a half million. So this is a large number relative to the populations in these countries. So apparently, there was a positive response to refugee crisis based on different sources. However, there was a challenge of policy responses and so in the long term, this seemed to be not adequate and there was a lot of resources and government in action and there seems to be an opinion that this has led to marginalization of refugees in some cases. So we are mostly doing quantitative research here, even though we're really interested in exploring the country context. So we tried to get some realistic ideas on what the government assistance to refugees in these countries looks like and all is mostly anecdotal based. It's very, very challenging to get even a basic idea on what the government assistance is, but all we know is that it doesn't seem to be enough to meet the basic needs. So in terms of, so our studies as of 2011 and that's basically because of data availability. So you would think that now, so just providing some background on what the numbers are in these countries and the composition of the refugees and IDPs. So the numbers for the first country, Armenia, are somewhat misleading in terms of the numbers because there has been some kind of change in the way refugees and IDPs are treated. So the real numbers are much larger than they are and so resettlement has resulted in the refugees not being counted as refugees anymore, but we have a reason to think that the refugees are nevertheless tangible sort of reality in this country. In Azerbaijan and Georgia, the share of refugees and IDPs are close to 7% of the population in all three cases, but particularly in Armenia. These are mostly ethnic Armenians that came to Armenia from Azerbaijan because of the war. Even though they share the ethnicity, in many cases there might be cultural differences as perceived by people. So to do this research, we use the Caucasus Barometer Survey that was conducted by Caucasus Research Resort Centers, has been used in other published studies. We, the reason why we look at 2011 way is because it provides data on preferences for refugee targeted assistance and as well as perceptions of refugee and other standard demographic socioeconomic variables that are important to consider when one looks at these kind of questions. So we restrict the sample to those who are in the age range of 21 to 65 and this is mostly to take out the students and retirees. We focus on those who are ethnic majority, they're not migrants and who drop those missing values and we end up with sample sizes as shown there. But if we define the age brackets a little bit differently, the results stay the same. So the variables of interest, the dependent variable is defined as a dummy, which takes one if the respondent agrees that the refugee assistance should be increased. 39% of population in Armenia agrees that their assistance should be increased. This is only 27 in Azerbaijan and surprisingly 73 in Georgia. But we try to understand the country context a little bit and we don't know yet. We are still exploring why this number is large. It probably helps to know that the independent variable of interest, the perception of refugees disadvantaged is also high in Georgia, so it's around 40% relative to just under 30% in Armenia and 20% in Azerbaijan. Whether people think that the refugees are integrated, so these numbers are very similar across the three countries and rather high around 80% in all countries. But to an extent, this is explained by the fact that most of the refugees share the ethnicity. So our controls in the baseline model include some demographic characteristics of individuals. So these are gender, their age cohort, the size of their household and the family status, whether people are married and whether they have children. We also take into account the socioeconomic characteristics of individuals, their education attainment, whether they're employed or not and the income of their household. And we take into consideration whether they reside in a capital, in other urban locality or whether they are from a rural area. So given how our dependent variable is defined, a dummy that takes one if the respondent supports government assistance to refugees, we estimate a basic probate model in the baseline approach. And so why is there a preference of redistribution? He is the perception of refugees, either whether people think that refugees are disadvantaged or whether people think refugees are integrated. We estimate models where these variables are included jointly, as well as where they are included one by one. The results are insensitive to the specification. But of course, a probate model would yield bias estimates because of unobserved heterogeneity. So people of all sorts of types that we can't really capture, they can be different. So people may respond positively to government assistance for reasons other than whether they think refugees are disadvantaged or not and we can't control for all of this in the model. So there is some unobserved heterogeneity. Of course, the standard approach to deal with this type of problem would be to estimate some sort of bivariate probate model, but this requires an exclusion of at least one reliable instrument and we were able to identify a good instrument in the data. So then we're not too discouraged with that. And what we do is, first of all, we try to control for as many characteristics of individuals as we can. So in extended specifications, we control for things that we think that are plausibly related to the independent variable as well as affecting the dependent variable. And we also do take some partial identification approach. And this is inspired by Altonji and it uses the amount of selections on unobservable as a guide to the amount of selection on observables. And the idea of this is to get some, to some measure of how large the selection on observables relative to selection on observables should be to take away the entire causal effect of the perceptions of refugees on government support. Support for government assistance. So I'm just going to now quickly talk through these results because the time is running out. So we find for a refugee disadvantage that this is a highly positive, highly significant correlated, so these are marginal effects reported here. Positive, highly significant correlate of people's decisions to support assistance in favor of refugees. For whether the refugees are integrated or not, we find different results in different places which is interesting. So in Armenia, we find that people's perceptions of whether refugees are integrated. If people think refugees are integrated into society, they're more likely to support refugee assistance. Whereas in Georgia, this variable is negative. Now we can think of different ways to explain this result. So if people interpret integration in terms of ethnic integration and they're likely to have group bias in group bias, then we would expect to observe what we see in Armenia. On the other hand, if we would think that the question of integration is perceived on economic grounds and whether people are employed and they are members of society and contribute to that, and that we can explain why there's a negative sign. So if you think that someone is well integrated in terms of their socioeconomic standing, that there's no reason to extend help in their favor. But again, I mean, our knowledge of country context is not sufficient at this stage to be able to convincingly argue why these different effects would exist, even though we can theoretically imagine that this could exist in principle. So quickly, just through this, in terms of additional controls, we included a number of characteristics that we thought could possibly capture some important sources of unassertive originating. So we generated measures of whether a person can be thought of as sexist or not, based on their attitudes to women's roles in the society. Some measures of religiosity, whether people favor competition or not, whether people think labor market is fair or not. We thought that by controlling to this, we can hopefully compare people that are more similar than what we had before. Controlling for these things, the results are the same. Now reporting the Altonji ratios. So again, these are ratios to indicate to what extent the selection and observables should be relative to selection and observables to entirely take away the causal effect of the perceptions that we are interested in. So here what we see, and the rule of thumb can be taken, the numbers that Altonji comes up with from his paper, and these are around 143.55 and he thinks that this is implausible. So by implausible means that it's unlikely that selection on observables is as large as that to take away the entire causal effect of the estimates that they're interested in. So therefore the conclusion would be that it is likely that some of the effect that we estimate is causal. So if we keep that in mind and look at these numbers here, we would see that the results that we obtain in terms of how refugee disadvantage relates to people's willingness to support assistance in favor of refugees, these numbers are rather high. Going back to whether people think refugees are integrated and how this affects people's willingness to support assistance in favor of refugees, we can see that only we can have some confidence about the numbers in Georgia that refugees insist the people's perception that refugees are integrated negatively relates with their willingness to support refugees. So summary, we find large positive effect of perceived disadvantage on willingness to increase government assistance to refugees. So some policy implications would follow in terms of raising refugee disadvantage in host country populations through media campaigns and so forth. Of course perceived and actual measures of disadvantage may diverge. And in many cases, there's anecdotal evidence that refugees are isolated and this may be important in shaping how refugees are in terms of their disadvantage and integration and so forth. And we don't find strong evidence that the role of perceived integration matters. But there are clear issues of measurement as well in terms of we don't know what the integration is perceived as, whether it is with respect to cultural integration, economic integration and we don't really know the, we don't have information on which refugees people have in mind when they talk about refugees. And there are obviously limitations so we only take, are able to do partial identification. We don't have an instrument to come up with robust causal inferences. It would be important to understand responses to different types of refugees, subject to data availability. An external validity of our study might be somewhat limited even though we motivated by saying that there are a large number of refugees around the world at the moment and important to understand, especially developing countries and important to understand the responses to refugees of this type. But we have to keep in mind that we're looking at a set settings, three settings where refugees are very close to the mainstream population in terms of their ethnicity and where large scale immigration is unusual. So we are cautious in suggesting the relevance of our studies for other country contexts. And so therefore it would be important to have some evidence from other countries as well. Thank you.