 I'm going to talk about a paper that we titled, Gone with the Wind and the National Migration. So there could be a lot behind that title. And this is joint work with our former student of mine, Emilia Abern, who's now at Victoria University in Wellington. So here are some numbers that we all know showing that migration is large, showing that migration is important, showing that we should care about migration, which I think in this audience is not something that I need to defend. Also, that migration will increase. And I put a conservative forecast here of the stock of migrants and about $400 million by 2050. Conservative because the point of the paper is trying to make that climate change is a big driving force. And the forecasts of climate migrants reach something from $200 million to $1 billion. So if you add that onto that, then the number gets larger. That's what this forecast is rather conservative. So what are we doing here in this paper? So I want to think about this in two steps. In the first step, what I want to do is I want to offer a joint analysis in the terms of economic driving forces of migration, climatic driving forces of migration, and political driving forces of migration. Combine those together. And the important thing is to have a combination of year-to-year variations of migration flows and of long-term, so a panel structure with large T through year-to-year variations. Then once we have this panel, once we run all our models and re-identify those variables that are key to driving or explaining migration flows, then in step two, we're going to plug them into the panel vector autoregressive model, where the question is, if you have a shock to one of those driving forces, how will migration respond over time? So the question will be, if I see a disaster today, how will migration respond over time? And again, just to reiterate, what is this joint analysis that I'm talking about? Well, this joint analysis is, again, the economic factors, right differences in income, economic opportunities, unemployment, and so on. Political, the usual controls will be warfare, terrorism, for that matter. We're going to add political freedom, the quality two measure that we heard about. And the climatic factors, disasters, weather and non-weather related disasters, and temperature trends. I'm not going to justify or defend that there is climate change. If you believe that there is no climate change, then probably this paper is not going to give you a lot. But the point is, we have climate change. We know that this will lead to or it leads to changes in the natural system. And then, for various channels, and this is just a subset of probably everything that goes on, this will have an effect on migration. And in this context, I'd like to think about migration as an adaptation strategy to this climate change. Now, what's the takeaway of this entire paper, if I don't make it through all the slides? Well, the takeaway is the time dimension, this year-to-year variation, is key to understand the effects of climate change. And if we look into the related literature, which normally uses decennial averages, decennial data, the results are either insignificant of whether related to temperatures or very small. And we find significant effects and very large effects. So this time dimension seems to be key. Climate change has, not in 20, 30, 50 years, large effects that already has large effects today, which leads, if you believe, the results to migration. And something that was surprising was that the effect of climate change, temperature here, and rather related disasters, is even more important than income and political freedom at the origin. So that's quite substantial and surprising. Effects of temperature are nonlinear and certain variables. And if we then look into dynamic effects of this shock, we find that the response varies quite a lot about the different categories of the shocks. A brief literature review, overview of the directly related literature, so this ignores single-country studies, is those five papers which mainly find small effects or a fine and parson's paper, hardly any significant effect at all. So before we estimate something, I'd just like to have theory, which explains what actually goes on, even if the model will be stylized, at least it gives us an idea and fixes our thoughts. So we have a Bohas-type model and the development of that model. And the idea is to come up with an equation, and there will be the bottom line here, that model's migration flows between a post-country I, a destination country J, over time. And that is basically explained by the differences in income and then the difference in this A variable, which is all kinds of social benefits. And those are related to political freedom, climate, and other socioeconomic factors. And then you have to subtract the moving cost from moving from I to J. And then there's a shock to it. Now if we then rewrite this theoretical equation, we have a gravity type equation that we can then estimate. And this alpha term here will capture the original fixed effects as in the related literature. And then the migration decision, again, is driven by the differences in the benefits in the destination country relative to the origin. Now there are a lot of econometric issues here. I'm not going to bore you with all the details. But what we end up doing is we have a baseline regression, which is all s. We're going to use the PPML estimator as a control. And even the negative binomial regression estimator is a further robustness, because we believe that it is actually the more appropriate estimator relative to the PPML. We're over this version and x is 0's in the data. What's the data set that we construct? Well, we end up constructing this bilateral panel data set, which is fairly large. So we have 16 destination countries. And we have the flow towards those 16 destination countries from about 200 countries. So this excludes basically really small countries, Holy See, for example. So I think it's pretty much the entire population. Period is 1980 to 2014. So that's 35 years. And those bold things over here just show you that the data set is very large. It's very large relative to the published or related literature and the time dimension and the panel dimension and NNT. What are the variables that we use? Well, the usual candidates, the usual suspects that you would like to have in this kind of augmented gravity model, migration costs, proxies by distance, and other socioeconomic dummies, borders, language alike, economic variables, income, proxy by GDP, share of young population as a measure of labor market chances, aid, agricultural land. And we'll talk about that later. Political variables, the war dummy, civil war, or war between countries, and the political framework the polity to measure. Climate variables, temperature anomalies on a country year basis, and disasters from MDOT database in a weather and a non-weather category. And that will be added to our regression model step by step. So let's get a feeling for what kind of data we're feeling here with. So this is a plot of disasters, total disasters in the sample over time. Blue will be weather-related disasters, and we see there's a clear upward trend. So in 1980, we had about 100 in total. Towards the end of the sample, it's more than 300. So tripled over time. For non-weather-related disasters, that's pretty much flat. There is a spike, and that spike is epidemics around the early 2000s. But that's basically it. And that's what we would expect in climate change, affecting weather-related disasters, but not earthquakes and the like. But besides a first-moment shift, there is also a higher-moment shift. And here, this probably poor attempt to depict this is the distribution in 1980 of disasters and 2014. And what we see is that this tail is just getting longer and longer and longer. So while in 1980, the highest a country could experience was 10 disasters. In 2014, it was more than 30. So it's not just increase in the mean. It is also increase in the distribution. So shift overall. Temperature, yep, we know that, is increasing over time. So 1980 was about the bait, the bench line here. And since then, it's basically upward trending over time, a bit of leveling up towards here. But 15 and 16 were record highs. So it's, again, starting to increase. Countries are differently affected. There is a small group of countries that experience small increases in temperature. And there is a larger group of countries that experienced larger by about 0.8 degree Celsius increases, quite a variation across countries. So let's look at the baseline model, very stylized, made up type model. And what we find is what basically the entire literature finds. So this shows that the stylized model is in line with different data sets. So for example, if we increase GDP at destination by 10%, migration would go up by about 10%. And this is robust across the all S with fixed effects. The negative binomial, the PPML, different kinds of fixed effects. So this is, we can take this to the bank, but this is a fairly robust finding. Now it does get more interesting if we add all those additional variables and go to what we call joint analysis, okay? So let's first think about what happens if we include wars and policy. Well, not much, only policy freedom at origin is significant. And this will increase migration flows. If we add dramatic factors, temperature and weather related disasters, we're here. Well, we migrate away from temperature into cooler countries and we migrate away from disasters. Now, what does this 0.03 mean? Well, that means that if temperature anomalies increase by 10%, then the migration would increase by about 3%, which is about 135,000 people per year. So that's roughly the dimension that we're talking about. And then, yes, young population as well is significant. So we move away from countries with high competition towards country with less competition in the lake. Now, you probably know, have a lot of questions in your mind. Is this really robust to whatever? Answer is, well, it is certainly robust to all of this. And we've done a lot of robustness checks here with all kinds of things that came to our mind and the referee's minds. And the results are very robust, okay? So you can't heal this temperature effect or the weather related disaster effect. Briefly, there are non-linear effects of climate change. So what we've done here is interacting temperature with the agricultural land share. So countries that rely more heavily on agriculture when they experience higher temperatures have stronger outward migration. Temperature and GDP. So you would expect that richer countries, countries with higher GDP should have better mitigation strategies and therefore have lower migration, that's what we find. And finally, the interaction of temperatures and weather related disasters is positive. So there is some correlation between the effects of climate change along temperature and weather related disaster dimension. And so we would expect more migration out of that country. Now, finally, going into the dynamic effects of those shocks, we now know which variables are significant and we're gonna look at four different shocks. Income at destination, the war at origin. Temperature at origin, disasters. Weather related disasters at origin. And this is gonna be a P bar X, so a panel vector auto regressive model with exogenous variables. And this entire thing rests with our identification assumptions. So we have four variables that should identify those four shocks. Unemployment should identify it to the P shock, okay? That's macro stylized literature. Epidemics might be an instrument for war. Volcanic activities in fact is an IV for temperature and my co-author actually has a DC in biology, so unfortunately she's not here, so I'll have to tell you how this works. Apparently, volcanic activity has a trade of between SO2 and CO2. SO2 rather cools with CO2 rather increases temperature, so the aggregate effect is not really clear in the outcome, but in any case, it should be a decent IV. And then finally, agricultural land should identify weather disaster shocks, okay? Through increased fertilizer usage and changes in the land use in the structure. So we plug that into the P bar, run it through and those are the impulse response points. So we have percentage change on vertical and years in the horizontal and we have about 90% confidence intervals here. If GDP at destination goes up, we have this expected increase over time, right? The destination country becomes richer and migrants starts to get in, so full factor. If there's a ward origin, we see this increase over time. For weather related disaster at origin, we have a bit of a surprise at least. To me at the beginning, there's nothing significant and then after about three, three and a half years, we have this increase in migration, which is fairly small relative to the other variables. But the most interesting part here is the effect of temperature shocks. And that is a decrease on impact and then overshooting, this would be an exchange rate, and overshooting of migration and a fairly persistent increase over time. So I wanna understand this one a bit more closely. And this can be done by combining two strands of literature really. So the first is Halliday, Piquet and Catania and Perry basically say, well they're binding liquidity constraints. If something bad happens, you can just immediately sell off all your assets and move away. So it takes time to liquidate your assets, make an informed decision through maybe networks, and then move out of the country. The second explanation by Dylan is spatial diversification. So you wanna send out members of your household in different geographic areas to ensure against those temperature shocks. Problem is with climate change, the risk probably goes up everywhere. So you have to send them out further. Sending people away further means higher migration costs. So you have to save for a longer period of time until you can actually make this migration decision happen. And this explains this initial lock in where migration first goes down and then over the saving, the relaxation of the liquidity constraint, the binding constraint, migration will increase. So wrapping up, again the two steps of this paper, joint analysis stressing the importance of the year-to-year variation and the long run effects to a large and large T type panel. And then estimating the dynamic effects of those shocks to migrations. And as we would expect, there's a complex mix of driving forces. Climate change as of today is an important driving force. More important, again, than incoming political freedom at origin. And those shocks have very persistent effects. There's a short window of opportunity where we can limit the damage that has been done. And that can be done by a combination of short run and long run policies, but that has been published a lot. And while finishing with this change of one of the most famous quotes of movie history, that is me. Thanks.