 I will move from a large region and a lot of hazards to a single country and to a single event. But in terms of geocoding and in terms of trying to map things, we're going to have some similarities. And also we will be trying here to use secondary data to quantify the impacts of disasters. Using this single event, which is called the Typhoon Damry as the example. So what we'll try to do is I'll try to see if we can quantify some effects of this typhoon that happened in Vietnam in 2005. This is the idea using the existing data. It's based on a paper written jointly with my former PhD and postdoc Lidang Chum, who's now an independent consultant. And if you want anything to do with geomapping or surveys, this is the guy you want to meet. I just wrote up some of the stuff but he did the hard work. And I'll show you the hard work later. Now, in terms of background, natural disasters may be increasing, maybe we report more natural disasters. But at least we have an increasing number reported of natural disasters in particular in coastal areas. And we would like to calculate the cost of these disasters. I do it partly because I can. As a researcher, there's an intrinsic interest in trying to do new stuff. But of course, if you have to do policy and policy advice, you want to show that there is some interesting results coming out of your research. For this case, we can think of insurance schemes and insurance policies because if we can quantify these impacts, we may be able to set up some insurance schemes for the farmers in this case. Cost-benefit analysis of course, socio-economic planning in these disaster prone or hazardous areas we could use for that. And then to learn more about these welfare implications of these disasters. So this is the main background. But as I said for me, it's because we can do it and it's good fun. Often when we have these disasters, we have two ways of measuring the cost of the disaster. First of all, whenever there is a hazard or resulting in a disaster, there will be disaster teams. And disaster teams will often do interviews with the people affected by the disasters and they will ask about the cost, about the impact of this, what happened to your family, what happened to your household, your fields. And of course, having people coming in at the time of the disaster, asking questions, people do not know the long-term results, but also it depends a lot on who's asking the question. And the Typhoon Damry is in Vietnam and it was in 2005, I was there. Some of the people coming to these disasters areas, that's the military. The Vietnamese army are moving out because they're very, very good at disaster preparedness and also the effects of these disasters. So if a member of the army is asking a farmer, so what happened, he or she may not give a... Well, the response will be affected by the fact that it's a military person asking the question. It could also be a foreign aid worker and then of course there could be a different response. So one thing about asking people directly during these disaster times is that the response depends a lot upon who's asking the question. The other thing you can do is, since we have regular household surveys in many, many developing countries, we can actually use the information in household surveys. Now, since household surveys are of course not implemented exactly at disaster time, it means that we have recall answers. And recall answers of course depend on perceptions and preparations. So the people's perception of the death or the impact of these disasters depends on whether or not they're used to these effects. So some people will be prone to many hazards. So they respond possibly with a lesser effect than people where the shock comes out of the open or out of the blue. So we know that for both of these surveys, answers will depend a lot on the individual situation. So what we're trying to do is use the household surveys, but then we will try to quantify it by comparison. And again, that is an impact effect. So we try to do this in an objective manner. That's what we do. To calculate this objective cost of the disaster, we need a very precise mapping of the disaster area, of course. And we need to have a precise mapping without asking people. That's why we use the scientists. Scientists do everything without asking anyone. So we will use some of their models. And then we will need a measure of welfare indicators in this disaster area. We have these from household surveys, standard household surveys. We will need a measure of welfare indicator in the disaster areas after the disaster without the disaster, which does not exist. But this is where we use all these comparisons. This is where the difference in difference methods come about and all these different comparisons of people. So this is truly an evaluation problem that we're used to in social sciences. This is at least how we try to set it up. Can we see the cost of a disaster as an impact evaluation? And we think we can. And that's what we try to show. So we try to have this feasible way of computing the objective cost using detailed storm data. I'll show these. Then we have the wind speed model from the scientists. The wind speed model will give us the disaster area without asking people. We have household survey data from this area. We have household survey data outside the area. We do a statistical comparison of the two to get some objective measure of the cost. And we try to answer questions who's affected by the typhoon we use this data. What are the short term impacts we compare with and without using household surveys? Are the impacts persistence, longer term impacts, again using the same data? And can we look into coping strategies from answers to the questionnaires? So this is what we try to answer. Getting to the storm or the typhoon or the hurricane that is caused in the U.S. I don't know if you know, but you can get reports as a typhoon or a tropical storm is emerging in any area in Southeast Asia. The Japanese meteorological station will give you early warnings. And they follow all of these typhoons or hurricanes. You can see that we have storm categories from a mild tropical depression to a tropical storm. Then we have the category one to five where it's truly serious. And you'll see from this mainly green area that this was a tropical storm. Then it hit the island, just, you know, the Chinese island and then moved into, this is Vietnam there, right? And our interest is the impact of what happened there. Okay, but we have the trajectory. So we have the center in some sense of the storm throughout and we have that for a lot of storms. Given this trajectory, we can actually calculate or we can try to compute an area. This is what happens when the storm hits the coastal area of Vietnam. And here again, we have the coast and you will see there are landslides, mudslides. Even though it's only a tropical storm, it has relatively severe effects. Housing went down and this is then after the storm where they tried to rebuild, right? Because of the flooding in the area, the coastal area. We're doing this. This is a wind speed model where you have the wind speed in a certain distance from the center of the storm. And this was developed by Holland in 1980. So you have the center and lots and lots of constants. This can be done in a spreadsheet. Okay, so if we have a dew mapping of the center of the storm, we can at any point, any distance from the storm, we can calculate a wind speed, right? So it's lower at the center than increasing rapidly and then decreasing. There are slight alterations because tropical storms have different angles, whether or not you're in the northern or the southern hemisphere, they turn. But that's just a matter of changing this a little bit, right? So it's a slight complication but still doable. So we know at these distances, we know wind speeds and then we say that you were hit by this damry storm if you were in a wind speed or in an area where the wind was above 35 knots. And 35 knots, I have no clue about that. But I looked it up, it's 40 miles per hour, 65 kilometers per hour in this situation, right? So that's a fairly, that's a high wind speed, right? It's okay. It will move people and move buildings, right? So what we can do is we can simply now we have this is the trajectory, this is the center of damry and then we have the upper or the northern and the southern parts where inside this region we had wind speeds above 35 knots, right? So we think of this, we define this now as the area affected by the hurricane damry or the tropical storm damry and areas outside were not affected in this situation. And all the black dots, these are actually, these are communes for which we have data, right? So we have data from the household surveys 2004, 2006, 2008 and all these black dot places we have a household data, a household surveys before and after the storm. So what we'll do is we will look at, we will try to find a comparison group for this and to find comparison groups we go back in time just as the flooding data and what we have is storms in Vietnam from 1951 to 2008 in this situation. I don't know if you can probably not see but in here, this is Vietnam, right? And each of these trajectories, that's a storm. So that's either a tropical depression or a tropical storm or a category one, the Red Star category one, two, three, four, five storms in this situation. We're using for all of these storms, we have a storm area, right? So we know that these places have been affected so we can compare these. So we will, we're trying to match, this is damry, the trajectory of damry and then we're asking whether or not they were storm prone or not, right? Because we tried to compare people, damry, this place, as you can see, it was no surprise as such. They were not prepared for storm of that speed at the present time but that was because of some forecast that went wrong but it's not the first time people living here see a tropical storm, right? So they protect themselves, of course, they take precautionary measures and sometimes they do but it also means that these are not random, you'll see much fewer storms in the southern parts of Vietnam, many more storms in the northern parts of Vietnam. So what we do is to select these areas, we use propensity score matching and multivariate match time, it doesn't really matter too much. The important variable is exactly the propensity to Oriental. The probability or the hazard of actually being affected by a storm and then we have the distance to the coast because wind speed is also taking off as you get inland, right? There are changes, also changes in the probability of mudslides as you move inland in this situation. Then we have the commune area, of course, if you have a small area you probably not hit as often and then the population size, how many people are affected by this. So we do matched sampling here and then we take before and after the storm to get some of the results. Our data, before damry, we have the Vietnam Household Living Standard Surveys pre-storm situation, then the storm happened in 2005. We have 2006 data for the same communes. So that's what we think of as short term. We are looking into, as I'll show you later, the paddy production, rice production in these areas and also housing. Then we have what we think of as longer term, three years, at least we may be able to see if they recover quickly in this situation. So we are going to compare 2004 to 2006 and 2004 to 2008. We have roughly 7,000 households in rural communes. Damry hit roughly 800 of these households in 264 communes of these surveys. So we select 801 households in these unaffected communes to form this comparison group. Then what we compare is their changes from 2004 to 2006. So we compare the changes in the unaffected households to the changes in the affected households and this is going to give us the impact evaluation. That's what we believe often. It's a fairly standard difference in different approach. So when we look at the affected and the unaffected area, you will see that the red dots are now the affected areas and then we compare mainly, as you can see, with a slightly northern part, southern part and then we actually have all the way down here but they will have lower weight in our comparison. So a lot of these. We have tried different cutoff points. It did not change our results very much. We tried to narrow it down in a very geographical, rigid way only focusing on this. The problem comparison here is that as we move north, we also move very much into rural areas and the Red River Delta and also where we have Hanoi close to this. So comparisons are affected by both of these. That's other changes not only being rural but also being closer and closer to Hanoi. We hope that this is not a major impact but we cannot test that. The outcome we look for is production of rice, patty rice. Then we look at the crop income and sidelight income to get an idea of the impact on their total income in the households. We look at food expenditure and we split this into self-consumed. That is consumption of own produce and then we look into what they bought in this situation and of course house repairs if possible. So these are the comparisons that we do. Short-term impact, looking at patty production. First of all, there was a yield loss, roughly 100 grams per square meter. That is actually 20% loss of the 2004 yield. So even though it doesn't appear as much in terms of weight, it's a fairly large relative share of this loss in patty production. Crop income, 30% lower but not statistically significant in this case. Sidelight income, no significant increase or decrease. So we cannot see them working more or less as such outside farming. Total income, we do not see a significant loss as you can see. No big change here but it is, crop income is lower. The point estimate is lower but it's not statistically significant. When you look at food, increase in self-consumed and we interpret this as there's been some of the livestock died. If the livestock died from the storm, you eat it. That means that you can increase self-consumption and there's a decrease then. There's a substitution between your own production, accidental production of meat and then what you buy in these situations. They have very large chicken production in some of these places, scavenging. So there isn't a substitution in this case. Total food consumption, we do see a decrease in the comparison. Turning to house repairs, there's an increase in the probability of having a house repair by 20% expenses increasing. We don't know but it's 25% relative to the poverty line. We have to have a fixed number here so we compare it to the poverty line and there's an increase in housing repair expenses. This is a very uncertain number what's reported here. So we mainly look at there's an increase so houses were affected in this situation. Longer term, 2008. No impact on paddy production so it's going down for a short while but then back up back up. They have two seasons there, no impact. Crop income, no change. Sightline income, no change. Total income, no change. Three years after this event, we don't see anything in their income or production. Self-consumed, there is a decrease in what they consume meaning that they had a stock, it died, so there's a decrease. So here is actually a capital depletion that has a longer term effect you can say. There's an increase again in food board but we don't see it as a significant increase and then no impact on total food consumption. They eat as much. House repair probability still an increase so it takes more time. Some people are probably postponing house repairs maybe because of pressure on housing repairs in the short term so the poorer, we don't know if they're poorer or richer the poorer may be postponing their house repairs and also we see a larger increase. So coping strategies as the final part households invest more in subsequent credit production. So they are repairing the land you would say, restoring the land. In terms of money or funding there's a flow of remittances on average for those who receive it's $100. There's an increase in borrowing, quite large increases in borrowing both formal and informal borrowing. The formal borrowing is mainly from state-owned banks and then disaster rate, not mentioned by many. So this short term disaster rate is not something they rely on in Vietnam. That's it. What we actually try to do is we try to develop this objective method to identify these storm areas hoping that it can be used not only in Vietnam but also elsewhere. We have mapping of wind and the center of the storms. We can use the wind speed model from the meteorologists to identify these storm-affected areas. If we have household surveys as we do have every two or every three years we may be able to use this difference-to-difference approach to get the impact. And this could be the foundation for these policy measures either private policy, cost-benefit analysis in terms of government interventions in these areas and then of course wider impacts of climate changes if possible. Thank you very much.