 talk about the relationship between natural disasters, poverty reduction, and remittances, and this is a joint work with a colleague from Ferdi, Alassan Drabot. It has been found in the literature that there is a negative relationship between natural disasters and economic growth, but also between natural disasters and poverty. However, there is much less evidence on the role of private mechanisms such as migrants transfer on poverty when natural disasters occur in poor countries. So we have few studies, these three studies that look at this relationship, but we would like to contribute much more and fill the gap in the literature. So we ask a simple research question, do private funds such as remittances help mitigate poverty in the context of natural disasters? Contrary to what other papers are doing, because other papers they take a sample of poor countries and they look at the impact of remittances after the natural disaster occur. What we are going to do is that we take the sample of low and low income countries. So these are very poor income countries. And we are going to look at the interaction between natural disasters and remittances on poverty variable, which is much more specific. And we are also looking at macro data. Most of the studies were conducted at the country level. So we have a panel of 52 developing countries over 27 years, 1984 to 2010. And also maybe a novelty in this paper is that instead of taking measure of disasters such as the number of people affected or the damage, which is kind of because these are determined by the poverty level of countries, we are going to use some exogenous measure of the intensity of disaster. And I'm going to talk about this later on. So we investigate the role of remittances in mitigating poverty in the context of disasters in a short-term perspective. So we use monetary poverty from the word development indicators as our main dependent variable. And we also try to deal with many of the endogenity issues that we have by using fixed effect models, some alternative estimation and a JMM model where we instrument our endogenous variable by their lack to values. So just to give you a flavor of the result, we found, indeed, a reducing effect of remittances on poverty in countries that experience natural disasters. But very interestingly, what we found is that actually remittances decrease even more poverty when natural disasters occur in countries. And I'm going to detail this later on. So we also run the analysis by looking at which type of disasters are driving our results. And we found out that the results are mainly driven by storms, hurricane, and extreme temperature events. So the literature, as I said, if we look at the relationship between natural disasters and poverty, it has been found for people belonging to the middle income, let's say, distribution, that disasters can push these people into poverty by destroying their assets, by eliminating their ability to rebuild their home, or just securing their basic needs. And if we look at the poorest, actually, disasters can even push them into a trap. So since poor countries, they generally live in very unfavorable conditions, they are already in a vulnerable situation. So disasters will increase their poor economic status. However, it's important to highlight that there are some heterogeneity of the effects of natural disaster on poverty, heterogeneity between the short term and the long term. And there is a very interesting paper by Ginu and Menendez. And what they do is that they look at the occurrence of earthquake in Indonesia since 1985. And interestingly, they find that after two years, so two years after the earthquake, indeed, the disaster would decrease the expenditure, the total expenditure. And then disaster has a negative effect on welfare of household. But then between two and six years, they found out that the effect of disasters started to be mitigated. And then from six to 12 years, they found that disasters has a positive effect on the welfare. And it's kind of very puzzling. But the way they explain it is that in the long term, the country has benefited from lots of public aid and good institution. And then the aid were very well managed. And then the negative effect, the primary negative effect of disasters, will turn out to be positive in the long term. So as I said, that's why we really pay attention here. And we only focus on the effect of disasters and remittances on poverty in the short term. So I mean, there is a large body of the literature that look at the impact of remittances. And in general, not only looking at remittances after natural disasters, it has been found that remittances, indeed, reduce poverty through accumulation of human capital and physical capital, also by reducing income inequalities and by increasing consumption. And when it comes to the specific case of disasters, it has been found that if households benefit from insurance mechanisms such as migrant transfer, this help them increase their level of resilience in the aftermath of shocks. Maybe here, because I thought I would not have time, but I can just elaborate a bit more. It has also, so most of these studies were conducted, as I said, at the country level. And there is one study that were conducted at the macro level. It's a study by Young, 2008, when he looks at rainfall shocks. And he found the same thing, basically. But this is the only macro study. So as I said, we use data from 52 low and low middle income countries. So this is the panel data. We look at the data from 1984 to 2010. We use two measures of poverty from World Bank Database. We look at the poverty-headcount ratio at $1.25 a day. And then we also look at, as a robustness check, the poverty gap at $2 a day. So our natural disaster variables are from the game data. So these were produced by Felbermeyer and Grosch. And what they do is that they use data from geophysics and climatology. And they assess the intensity, what we can call exogenous measure of intensity of disaster. Exogenous, because this is not like the number of people who were injured, or who died, or the damage cost. And I'll talk about this when I will discuss the $190. This is impacted by the poverty level of countries. So we use first the aggregated disaster index and then the disaggregated measure. So we have, first, the wind speed. So this is the maximum wind speed in knots for storms and hurricane. And then we use a measure of temperature, I haven't mentioned, but a measure of temperature, which is the difference between the maximum monthly temperature and the mean of the period. We also use a measure for drought. So it's a dummy, equal to one. If in a country there is three successive months or five months in a year where the rainfall level is below 50% compared to the mean of the period, there is a measure for earthquake. So it's the maximum on the risk scale. And we also use the maximum of volcanic explosivity index for volcanic eruptions. So our imitants variable represent the transfer received in the countries over the period. And we control for various country characteristics. So we control for the quality of the institution to a measure for democratic or autocratic countries. We also control for population variable, total population, population density. We control for urbanization rate, which also capture internal migration. And we control for the growth rate of real GDP per capita, which capture economic factors such as unemployment and quantity and quality of infrastructure. So our model, we run a country fixed effect model where our main interest variable is the interaction term between natural disaster and remittances. We also add that disaster and remittances separately. And we add our control variables with one year lacked, because we assume that all of these variables are very endogenous. And it will be better to take one year lack, which makes them a bit more exogenous. Because they are affected by poverty as well. So we control for time invariant country characteristics, which are unobservable by using a country fixed effect. And we add also some time dummies. About the endogeneity, we have many issues of endogeneity. So as I mentioned in the beginning, the first one is the potential measurement error of the number of intensity or intensity of natural disaster due to misreporting or maybe sometimes government can inflate numbers for it purpose and so on. Also, the intensity of natural disasters, as I said, if you have an earthquake in Haiti with the same level of intensity and in Japan, you would not necessarily have the same number of people affected. And this is due to the poverty level of the country itself. So the intensity of natural disasters may be influenced by the level of poverty. And then that's why using an exogenous measure of disaster is very important in here. And maybe more seriously, it's the endogeneity of remittances. Ideally, we would have an instrumental variable which would affect poverty only through its effect on remittances, but we have not find it. So if you have ideas, they are very welcome. So what we do is that we try to manage this and try to run a series of robustness tests. So just to explain maybe why we have an endogeneity issue, first of all, there is likely a reverse causality. So the amount of remittances received can also be explained by the level of poverty of a country. That's the first thing. Second, poverty determines the location or even the migration choice. And then the future receptor of remittances. It's very likely that poor people live in areas prone to disasters, for example. Or maybe they are too poor, so they cannot afford migration costs. And even if this area is very prone to disaster, they might want to live, but they cannot. So we have to deal with these kind of issues. So the solution we find to mitigate the absence of instruments is that first we consider our measure of remittances, the logarithm of remittances received one year before with a T minus one. So with one year lacked. Instead of using a contemporaneous measure of remittances because it's very likely that the remittances that you receive in T minus one will affect your poverty in T, but the reverse is very unlike. That's the first thing. The second thing is that we run a JMM model and we instrument our endogenous variables by their lacked values. We also control in our regression for time-fixed effect, I said it, and we use disasters and remittances not only for T, but T minus one in the same regression. So to capture some kind of ex ante effect of remittances in helping household build resilience. I don't know if you see something. So it's pretty small. In the first two columns, we just run the plain regression. So the first column, we just look at the correlation between the interaction between remittances and disasters and disaster and remittances without country-fixed effect, without time-fixed effect, just a random effect. And we find out that the interaction term is negative and significant, meaning that indeed, when a disaster occurs, remittances help mitigate its effect on poverty. Disaster is positive, suggesting that it increased the poverty level of countries and remittances is negative. So we add our control variables and our results are robust to the addition of these control variables. And then we run the country-fixed effect. So we control now for an observable time in variance-country characteristics in column four and the results hold. We add the controls and our benchmark is the last regression, the column five, where we control for both country-fixed effect and time-fixed effect. And you can see that the interaction term between remittances and disaster is still negative and significant. Disaster has the positive and significant sign. However, although remittances is negative, it lost its significance. So we have to be careful here because it doesn't mean that remittances do not have an effect anymore on poverty level. However, because this non-significant of remittances should be interpreted with the interaction term. And it just means that indeed when disaster or cool actually the impact of remittances on poverty is even stronger just to interpret the magnitude. So for countries experiencing an increase in the disaster index by 1% and receiving the average logarithm of remittances in the sample, the poverty-head count ratio at $1.25 a day is expected to decrease by 1.15. And just we take the coefficient associated with the disaster index minus the coefficient associated with the interaction term times the average of logarithm remittances. So now we desegregate the disaster index and we look at what drives our result, right? So we look at disaster by disaster. So you can find, you can see that for all of our disaster the interaction term is negative. However, the effect is only significant for wind speed. So storms and hurricane, the measure of storms and hurricane and temperature. So the difference in extreme temperature events and drought. Suggesting that our results are driven by these three types of disasters. We start the series of robustness check where we control in our regression both for remittances and disasters, not only at T, but at T minus one with one year lag. And we have the results, sorry, are still robust and they are mainly driven by wind speed, temperature and drought. However, when we, so we don't use any more remittances at T, but the lag value of remittances which is supposed to be more exogenous than remittances measure at T, we find out that our result for the aggregated measure still holds for wind speed and temperature. However, drought lost its significance. So if we want to be more conservative we would say that our results are mainly driven by storms and hurricanes and extreme temperature events. So here we just run the JMM estimation model and again it's the same and drought is not significant either. And we change here the dependent variable. Instead of using the poverty head count ratio we use the poverty gap at $2 a day which measures more the incidence and the depth of poverty and we pretty much found the same results as well. So what can we take from this? So private funds indeed such as remittances reduce poverty in the context of disasters but interestingly remittances also have an exante role by helping household build their resilience to shock. Social networks and migrants in particular are very important channels for countries which can use them to deal with adverse effects of shock and just to compare with the results by Ginu and Menendez which showed that public aid could indeed help mitigate and even produce some positive effect on welfare private funds can be used immediately after the shock. Public aid they take time to be organized and so on and may reach communities much later on whereas when it comes to remittances or migrant surfer or private insurance mechanism when a shock happened you can just send an SMS or a call and say I have there is an adverse shock here I need some help and assistance. However, and I haven't mentioned this but it's very important and that would be the follow up to look at also effect on inequalities because what happened to people who do not receive remittances and who live maybe in very poor countries where governments cannot really help or the help would be limited because we know that maybe most of the migrants belong to the middle income distribution and for these people, so the receptor for remittances can be positive but it can also have some negative effect on inequalities for people who do not receive it. So we may want also to look at that in future research and finally, so it's also very important to be able to combine both private mechanism and public mechanism in the aftermath of shocks and of course, I mean, this has been said again and again but I repeated here that would be also important to reduce the cost of sending remittances. It will overall have an impact on the amount that people will receive after the shock. Yeah, thank you.