 First of all, I think it's a pleasure to be here and a pleasure to open this session. As you heard during the introduction, I'm an economist by background. I will try to avoid using economic terms too much to refer to people as I realize that this isn't always the optimal way to discuss such matters. This is fundamentally taking an economic idea of migration and trying to see how we can use this, or if we can use it as maybe a more accurate way to say this, to see if we can use this to understand the patterns and dynamics of the refugee crisis that Europe is currently facing. This is a single paper. We've been working on it for a little while, but the results are quite preliminary. As I might talk a little bit about later, the data requirements and the comparability of sources in particular aren't as good as we would like, and this has taken quite a lot of work to get to where we are. So we're using data that we know is imperfect, and we want to try and understand more specifically about trends and patterns rather than exact numbers, and that's how we'll try and present this today. I'll try not to go off into too many Greek letters and too many equations. I think we have a very interdisciplinary audience, and I don't want to turn everyone off with a whole raft of equations. But if you see an equation, don't get too bored. I will try and explain what I mean. So a bit of an outline of the presentation. I've split it into seven parts. One to six are quite typical motivation. A model of migration that we try and augment to understand what the push factors of refugees and what the pull factors of the selection of destination countries might be. We then discuss the empirical methods. I'll try not to go into too much detail here. And then we have a little bit of discussion about the multiple data sources that we try and pull together. Discuss quickly the results and the conclusions. And then in the final section, I want to discuss the next steps, which in terms of application is actually probably the most important part of the work yet. Unfortunately, it's also the part that we actually haven't got around to yet. So as I said, it's an economic model and in economics, migration tends to focus on what are called push and pull factors of migration. So a push factor are things that cause people to want to leave their country of origin. So these could be things like adverse economic circumstances, lack of opportunity and so forth. What we're going to add to this is that features such as fragility or conflict or repression might also be able to be added to this model and might be able to provide some understanding of why people would like to leave. Pull factors are factors in the destination country. So if you choose to migrate, then you may well face a choice of multiple potential destination countries. And so it's not just the decision to leave, but also the decision of where to go that becomes important. And again, these can relate to things like public services, of course, the economic situation, employment, education and so forth. And again, what we want to do in this situation is try and augment this model by adding variables such as German von Kornenskultur, for example, that may actually influence a decision of someone to select Germany over another country in Europe if they're choosing to or attempting to migrate. Essentially, these models all boil down to the same thing, however. And that's that the decision to migrate is based on the net present value of migration. So if you're going to, in economic terms, I do apologize, but if your utility is greater from moving than from staying, then you're going to move. Although, of course, this is going to be taken across every single interval. So it's not simply enough to say, right now, if my trade-off of my expected benefits of migration are greater than my expected costs that I should move, but rather it is the dynamics of that that are important, that my expected benefit in every subsequent period after I move is better than my subsequent benefit of staying. The reason why we want to do this is... Now, I'm taking words here that others use. Basically, it's impossible to look at the news at the moment without reading words like migrant crisis. And what we want to do is to try and understand precisely what the dynamics of this movement are. Why are people leaving? Why are they coming to the countries that they choose to come to? I mean, I think the Syrian Civil War is what seems to dominate most of the discussions, but there's a lot going on out there. Repression in Eritrea is an example. Iraq, Afghanistan, Libya, and so forth. And so what we wanted to do, we set out basically to test whether or not an economic model could be used to understand these patterns of migration as well as a kind of what we would call voluntary migration. So low voluntary migration may well be based on largely rational ideas. It's possible that at least on the push side, we would find something different with involuntary migration, but that's essentially the hypothesis that we set out to test. What we want to do is understand what the push factors of the crisis are. So to see which events throughout the world have caused migration from country A to country B. We also want to understand if the pull factors are a significant driver of this. And then finally, we also want to understand if or how the crisis itself might actually wind down, what might happen in the future or what could happen in the future that would change these patterns both on the push side and on the pull side. When one looks at traditional economic studies of migration, one sees that basically the most important thing is the relative economic state. So people will have a tendency to move from poorer countries to richer countries, from countries that are stagnant to countries that are growing, from countries where incomes are low to countries where incomes are high, and from countries where they struggle to get a job to countries where they can get a job. However, a recent more recent work, and I say recent in a sort of relative sense because we're talking about models that go back to the 50s here. And then in the 70s, people came along and they added things like the quality and availability of public services as a pull factor. And then what's called partial adjustment and network effects. So network effects are essentially the network that you're going to have when you arrive in a new country. So if there are lots and lots of migrants that have already made the journey from your home country to a destination country, then that might make it easier to integrate, easier to get a job, and so forth. And so this is going to be important. And partial adjustment is essentially saying that people have expectations of what they're going to have when they migrate, that aren't necessarily based on the present, but rather on stories they've heard from people that have migrated previously. So if you know a family member that has previously migrated, who's been able to integrate and get a job and so forth, then that is going to lead to adjustments of your own expectations. Whether or not they actually match the reality at time to your not here is rather irrelevant. Then there are other things like geographic and cultural closeness, and so forth. In the case of forced migration, what we're saying is that these could be augmented by circumstances in the source countries, such as conflict, repression, and so forth. And specifically when we're considering Europe here, as you can see by what we said here, there are things that can also influence this at the destination countries. So we want to understand if Germany, if Angela Merkel essentially suspending the Dublin Convention and refugees had an impact, if the EU-Turkey deal has actually had an impact, if the changes in the capacities and role of front-tacks had an impact, and so forth. So here's where we get to the equations. Had to end in 1995, although this is a sort of a recent manifestation of a long line of literature, basically said that migration is a decision of a utility maximizing individual. The probability of migration then depends on the expected utility at the origin and the destination countries. That's all that this equation says. Basically, if we have a... So we have your expected utility at the destination country, manage your expected utility by staying, plus personal influences, costs of migration, and so forth. So fundamentally, if this equation is positive, then economics posits that an individual will choose to migrate. Borjas in 1987 extended the basic framework that we're discussing here to include a probability of employment in public services, and so forth. So basically we can rewrite this previous equation like this. So what we're saying here is that the difference in your utility from moving is wages in the destination country minus wages in the origin country, difference in public services, and the differences in the probability of employment. Because this is a logarithmic equation, the probability of gaining employment multiplied by wages becomes additive. It's that simple. Our postulation, just to formalize what I said before, is that this can be further augmented to include the push and pull factors of forced migration. And so what we include here are the pull factors at the destination country and the pull factors in the origin country. Otherwise, this is essentially exactly the same model, and this is the hypothesis that we're telling, that we're testing. As I said, it's important to note that migration is dynamic. That your decision at time t is not just based on your utility at time t, but also your subsequent utility in t plus one, t plus two, t plus three, and so forth. At the same time, because of network effects and so on, your decision at time t is also influenced by other decisions at time t minus one, t minus two, and so forth. And this allows us to write aggregate migration as this equation. Where basically the amount of migration is related to the number of people who are better off by migrating than not. The theoretical predictions of this model or of our extended model in equation three, I should say, is that worsening circumstances in an origin country will increase migration to all destinations. And so this could be a worsening economic climate, but in our hypothesis, it could also be worsening repression. It could also be worsening conflict. It could also be longation of conflict. It could be international actors entering into a conflict and causing further confusion. And at the same goes that if the situations improve, then obviously this should lead to reductions in net migration. Policies at the destination country can also have similar impacts. And so a policy in a single destination country should, according to this model, increase inward migration to that destination from all origins. Now, obviously, if every single country in the sample, sorry, decides on exactly the same policy, then we're going to find that migration to every single destination country should also increase. Okay. Basically, as migration is dynamic, I'm going to go back to this equation and do apologize. But this highlighted in red is the most important part. Because what we're saying here is that it's not just a case of your circumstances at time T, but also how you expect your circumstances in the future to be. So it's not just a case of saying Syria faces a conflict or people in Eritrea face repression, but rather the expectations that those people are going to have in the future. And we know from other conflict research how complicated this can be because exposure to conflict itself has very damaging impact on individuals' expectations of the future, which can then lead to adverse selection out of education or adverse selection out of the labor market, out of entrepreneurship, and so forth. Anyway, with some clever messing around with the equations, we can basically get the following econometric specification. And what we're particularly interested here are the dynamics of migration. So we have the lagged migration, migration from the previous period, and the migrant stock. So the sum of all migrants that have come from a country to a sourced country. We're also interested in, like, control variables and the changing control variables when this is available to us. And so what we're trying to estimate is migration from an origin, sorry, to a destination country, from an origin country, which is based on basically unique aspects of relationships between countries. So if you take two countries, their history of migration, their cultural closeness, and so forth, are likely to be fixed in time, and that's what this captures. Similarly, there's likely to be something that relates to the specific time period. So whether this is to do with policies themselves, whether this is to do with the push factors or something like that, there are things that are going to be unique to the time period. And we control for both of those. We control for past migration and the migrant stock. And then finally, we're going to look at these control variables. And what we do with these control variables, as I said, is we take typical economic migration models, we control for everything that we can that is usually significant in those, and then we add these push and pull factors that are related to policies and to situations around the world. Okay, if the terms 2D panel and 3D panel mean nothing to you, I don't blame you. I didn't know what they meant until I started this paper, either. But traditional migration literature looks at typically either all inflows to a single country or inflows from multiple countries into a single country. And what has often been missing is that actually this doesn't capture everything. That the relationship between Afghanistan and Germany is probably going to be different to the relationship between Turkey and Germany and so forth. That there are unique impacts there that are driving those levels of migration. And so what we want to do is use a so-called 3D panel which was sent at all in 2012 or some of the recent people to use. Others have done it as well. And what this does is it creates country dyads. So you look at the relationship of migration from every destination country, from every origin country to every destination country. And so what we did is we created dyads between the EU 28 countries in the EU and five illustrative origin countries. We're going to expand this to include a lot more information as and when we can. But for the moment, because of the extensive data requirements and so forth, this has been a little bit too difficult. So we focus on these five countries just for now. We use a bunch of panel estimators. If you don't know about panel estimators, that's okay, don't worry about it. If you do, we want to correct for dynamic panel biases. So in addition to the typical FE, which we know is going to be biased, we also run the Arlano bond first difference GMM, Arlano bond system GMM and the Pesaran mean group CCD. Data requirements are quite significant because we need data on dyadic migration, economic data for origins, which is sorry for destinations, which is easy enough. But when we're dealing with conflict effect countries, repressed states with a lot of repression and so forth, it's a lot more difficult. So we have to do a bit of legwork there. We look at violence, fragility, repression and other political data in origin countries and policy data in that, that should say destination country, sorry. So what we use is imperfect. We know this is going to be imperfect because we don't, but we don't have anything better. So we use the first time asylum applications by origin and destination country from UNHCR. This isn't going to give us to have an exact number, but it will allow us to discuss trends and so forth. Economic data from the origin countries is badly pieced together from the World Bank, CAA source book and various estimations. We use quarterly data. Often all we can get is annual data. So we use all of the quarterly data we can to estimate what the quarter figure from the annual figure should be. For the destinations, it's easy enough. We get it from EuroSTAT, violence. We use UCDP where we have it. We use ACLED where there's no UCDP. We use UCDP where there's no ACLED. We use journalistic sources and so forth. And policy data we take from journalistic sources. So we include current migration, lag migration, and moving total of lag migration. And perhaps most importantly, the probability that an individual will be allowed to stay if they reach the destination country. We include socioeconomic variables. We include the conflict event counts and major political upheavals. We use changes in the EU border force. The de facto changes to the Dublin Convention and so forth. We also use the inverse distance between the two capital cities for each dyad to attempt to, at least in some way, control for the cost of migration. I'll skip this for now. Okay, so what we find is, to those that don't know, stars are good when you're looking for a statistical significant finding. And we have a lot of stars in this graph. In fact, we have the maximum number of stars we could have, which is really good because it matches our priors. So what we're saying here is that there is a strong positive predicting effect of current migration based on migration. And so we have this partial correction that we were discussing. We have a large positive and significant effect from the migrant stock. So the more people that come, the more people that are in a country, the more people will come. And reassuringly, the probability of an individual getting an individual's asylum application being successful is a strong predictor. Socioeconomic variables, as you can see, we're losing stars. In a way, as an economist, this is perhaps a bad thing, but it's very understandable in the particular circumstances that we're dealing with. That actually what we're seeing or what we're finding or what we're seeing here is that there is no economic effect, is not economics that is driving the migration itself. And finally, when we look at the origin and destination variables, what we find is that really big events. So in Iraq, the period after Islamic State taking Mosul, for example, is a strong and significant predictor of migration from Iraq to all European countries. Similarly, the duration of the Syrian conflict is also positive and significant. We don't find the same for Afghanistan, Libya and Eritrea. And we also don't find it for small, relatively small events, so conflict event counts and so forth, we don't find it for. So this is perhaps quite surprising in a number of ways. Because what we see is that there are certainly drivers at the origin countries, but not at the source countries, or sorry, not at the destination countries. And we think what we're finding here is that because most policies that have been enacted, essentially apply to the entirety of the EU, that we don't see a differential effect as a result of that. Perhaps more reassuringly though, we also show that barriers, so these are when countries have erected physical barriers, they have absolutely no impact at all. So at least this is maybe something that we can take away and something we can recommend, that the barriers don't work so the state may not save money and look a little better. So what we would conclude here is that in part we have some information from economics that we can use in this crisis, to understand this crisis, but that not everything that predicts voluntary migration, shall we say, is something that also predicts involuntary migration. And I think that's a very important takeaway because we're saying, yes, we can see some of the impact of repression, of conflict, of violence, of fragility, and what is driving people, but we don't see much on the country side, on the destination country side. And here, that's something that, as I said, I think is important to take away. That fundamentally, it says that any policy that aims to tackle this has to tackle the push factors and not the pull factors. For the next steps, it's all fine and well having stars and nice tables and everything like that, but what we really need to do is do some other sample predictions. So this would allow us to test a range of hypothesis about forced migration and would also allow us to try and understand, not to predict the future per se, but to understand what would happen in certain hypothetical situations in the future. The first thing that this allows us to do is to test the accuracy of the coefficients, that we have. So what we want to do is see if, actually, if we run this in a subset from 2008 to 2014, how well do those coefficients predict what happened in 2015, where we also have the data? And it will also allow us to test future hypothesis by testing the impacts of changes and key variables. So if we say, well, actually, if the Syrians have a word where to end tomorrow, what would the impact of this be? We also want to repeat the entire process against previous migration crisis, which would allow us to be a lot more accurate and, again, a lot more certain in what we're saying here, to understand if these crises wind down in the way that we predict. And so what we want to do here is basically do everything we've done here and these other sample predictions using data, which will hopefully allow us to, as I said, be a little more certain about the strength of what we're finding. Thank you.