 One of the central motivations of my PhD is that I've seen a lot of studies both in economics and in other social sciences that try to explain or predict the impact of a disaster or a risk without really taking into account the different components of a disaster or a risk. So this leads to, in my view, it might lead to some problems with indigeneity, it might lead to some problems with reverse causality, which are definitely problematic. So there is, when you read reports from the UN or the World Bank on disaster risk, it seems to be the case that most people agree that there are these different components when you talk about risk or disasters. There needs to be a hazard. A hazard could be, for instance, a natural event such as a drought or a flood. It might also be a disease or also human-created hazards, technological hazards such as a nuclear power plant that might explode. Then for it to be a real risk or a disaster, there also needs to be some exposure. If an earthquake happens in the middle of Antarctica, nobody cares and it will not turn into a disaster. So there needs to be somebody exposed to this hazard. Thirdly, those exposed also need to be vulnerable to this hazard. So this is why, for instance, I think in 2014 was it the great earthquake of Haiti happened in 2010, yeah, don't look at me. There was a great earthquake in Haiti the same year. There was a similar magnitude earthquake in Chile, but we didn't hear as much about the earthquake in Chile because people living in those exposed to that earthquake were clearly not as vulnerable as those living in Haiti. So we have these three components and when you just look at disasters as such and try to see what are the impacts of this disaster, it's important to separate these three components because if you don't, then you might, what you measure when you look at a disaster might be the vulnerability, for instance, which is definitely correlated with what you might want to explain. So, for instance, if you want to look at the economic impact of a disaster, then you cannot just look at the disaster as such. You need to separate the, filter out the exogenous component of this and the hazard component is definitely the most exogenous. I'm not saying that in all cases it is exogenous, but in some cases it is. And I think Henrik is going to come back to a study where he uses this exogenous component of the disaster risk in order to look at the impacts. So this is one of my favorite figures in what I've read about disasters. This is not mine. This is from paper by David Strumper from 2007. And he just plots all the disasters that are available in a global database for disasters, which is called the Emergency Database or the EMDAT. And here we see a stylized fact that is often cited, especially in the introductions to almost all papers about disasters, is that there is an increased frequency and magnitude of disasters in the world. And we clearly see that the number of disasters over this period from 1960 to 2005 is growing, almost exponentially growing. So why is this the case? So this might be the case because either the hazard component of this disaster is growing. It might be that Mother Earth has just become more angry at us and creates more storms and more droughts, more earthquakes. It might also be because the exposure component is growing that people tend to, or that firstly, that the population is growing and spreading out to more areas and also that people may tend to locate themselves in hazardous environments such as coasts or drought-prone areas. Might also be the case that simply the vulnerability is increasing, that modern constructions are not as safe as traditional constructions. Although this seems not to be the case since the total number of killed in all these disasters is not growing, and also the number of killed per disaster is decreasing actually. And this last fact here leads me to think that the main reason we see this increasing in the number of disasters is probably because of more and more complete reporting, simply that more events enter into this database as the years go by because people are more aware of the fact that there are disastrous events taking place. Anyway, here in this study what we want to look at is the exposure component. So our question is how is the exposure to risk changed since 2000 in Sub-Saharan Africa? And we look at the period 2000 to 2015. And we completely disregard the vulnerability component, so we look at hazards and we look at population, and then we overlay the two and we get an idea about how exposure to different risks have evolved in Sub-Saharan Africa. There are a few so-called megatrends that take place in Sub-Saharan Africa. We have population growth. The continent that sees the largest population growth these years. There's a lot of urbanization going on. Coupled with this, we also know that the climate is changing. We have global warming, and this is particularly true for many places in Sub-Saharan Africa. But then on the side of this, we also observe economic development. Most of the countries in Africa see high growth rates. So this is definitely also something to take into account. Whereas the population growth is clearly something that will increase exposure to hazards. Climate change might increase the hazard component, whereas economic development might reduce the vulnerability to these hazards. Okay, so which are the risk factors we look at here? And this might seem a little bit arbitrary that we choose these four risk factors, but we wanted to cover sort of different areas when you look at risks. So we are looking at the risk of being exposed to drought and to floods. And then we look at malaria. We wanted to include a disease component, and malaria is an important disease in Sub-Saharan Africa. We might as well have chosen a different disease, but this was the one that we could obtain data for the whole continent for. And then we also look at the risk of being exposed to conflict or political violence. And I'm going to now take you through each of these risk types to explain how we measure these hazards. And then what we do is we make sort of hazard maps for these four different risk types. And then we overlay these hazard maps with population maps. Okay, so this is basically it. It's very simple in a way. We just divide Africa into tiny pixels of one time one kilometer. And then we have an idea of how many people live in each of these pixels. And we also have an idea of what is the risk of each of these types of hazards to occur in each of those pixels. So it's simply a geographical overlay analysis, I would say. It's not very economics. This is more geography what we're doing here. So the population data that we are using is available for the years 2000 and 2005, 2010 and 2015. So we have four observations here. To give you an idea about the number of observations that we are working with, there are 29 million pixels each year. So we have a lot of observations at a very detailed level. Then in order to separate rural areas and urban areas, we define an urban area in a given year. If a specific pixel has a population density of more than 150 per square kilometer, or in this case it's just a pixel has more than 150 people living in it, and it also has a travel distance which is lower than one hour to the nearest urban center of at least 50,000 inhabitants. This is what the population data looks like. So this is just a population density map of sub-Saharan Africa. And this data comes from the World Pup Project, which is based at Oxford, I think. They use census data, the most recent census data for all these countries at the sub-national level. And then they couple it with GIS information on all kinds of things, roads and buildings and nighttime lights, to spread out the population so that it sort of estimates more or less correctly at the very detailed level. Of course this could be done better, I guess, if you employ more data and more satellite images, for instance. But as of now, this is one of the best data bases that I could get hold of, at least. If we then move on to the drought hazard. So here we are working with a drought index that I have worked a lot on in a separate paper in my PhD. Essentially this drought index captures a combination of rainfall, temperature and greenness as observed by satellites and the anomalies in these. So if you're interested, then the way I constructed the index was to look at anomalies in greenness and try to predict these anomalies and greenness by anomalies in rainfall and temperature. What you see on this map is the number of years during this period, 2014, in which there had been a drought in each pixel. And the way we define a drought is that at least half the months of the growing period in each pixel there is a situation where this predicted greenness index that we create is below a certain threshold. And in this case, the threshold is the 10th percentile of the overall distribution of the drought. It's a little bit complicated to explain, but you get an idea that the areas that are sort of green or yellow here are the areas where there's a lot of variability in rainfall, where there's a big risk that in a certain year rainfall and thus greenness will go below a certain threshold. That means that it's very difficult to grow agricultural products. And we see that there's a lot of hazard in the Sahel belt and in East Africa around the Horn of Africa, but also in the southern part, whereas in those areas where we see generally lush in green conditions, there's not the risk of going below the threshold during the growing period is not as big. Okay, the flood data is not something we have created ourselves. This is something we just derived from a UN database. The UN Agency for Environmental Protection has this online portal where they have developed a very neat database also at the right resolution for us, which was these one kilometer pixels. It only contains the risk of riverine floods. So no coastal floods are in this database, but we found that only less than 20% of the floods in Sub-Saharan Africa were coastal. So we are looking at riverine floods here and it's a combination of many different methods that will not go too much into detail. But what this map shows is the number of flood events that will occur in each pixel for each 100 years. And the flood events here are the ones that where a river turns into a flood and there will be surface water in a big area around the flood. So we don't look at excess rainfall in places that are not close to rivers as such. So we look at these floods where a river goes beyond its borders and floods a big area. Okay, malaria. This is also something we have drawn from a different source. This is the Malaria Atlas project. The same people who created the population data, they have a big database on malaria. They recently or maybe almost a year ago actually published a paper in Nature where they showed that the prevalence of malaria has declined dramatically in Sub-Saharan Africa in this period from 2000 to 2015. Actually the prevalence which is defined here as the share of zero to two-year-olds who have malaria has dropped from 33% in 2000 to 16% in 2015. So we use the data that they created for this study as well. Conflict risk. This is a little bit more tricky and I show you here a very colorful map which may not say so much, I don't know. But what this shows is for the whole period all the conflict events that are contained in the ACLID database, armed conflict location events database, which I'm sure many of you are familiar with. It contains the GPS coordinates for all conflict events or political violence events. And what this map shows is sort of an attempt to calculate the predicted number of fatalities for each of these pixels. So what we do is with departure in each of the locations of each of these events sort of spread out the number of fatalities in a given radius using kernel density smoothing. And then this is one of the measures we use as conflict exposure. We also have a more simple, more broader definition which is simply if you are located within 25 kilometers of a conflict event with at least 10 victims, then you're exposed and if you're not located within that in a given year, then you're not exposed. So we have these two definitions of conflict because this one yields a very narrow definition which is actually the risk of being killed in conflict which is still relatively small, whereas being exposed to conflict there might be other consequences that being killed from being near a conflict. So we use these two definitions, one narrow and one broad. So now I'm getting to the results part of the town. What this graph shows you is at different probability levels, what is the share of the population in Sub-Saharan Africa that is exposed to each of these types of risks. So if we look at the far end here, the 5% risk level which is relatively low probability, we see that a lot of people, a large share of the population is exposed to drought, relatively large share of the population is exposed to malaria and conflict using the broad definition that I explained before being within 25 kilometers away from a serious event. These probabilities are calculated, you know, counting the number of times in this period this has happened and then dividing by the number of years. So it's also a pretty crude definition of probability but still it is something, whereas the higher probabilities we see lower shares. And generally what is also evident here is that a lot of people are exposed to droughts, not so many people are exposed to floods. But remember that this definition of a flood is to be completely covered in water close to a river, so it's not the risk of having excess rainfall on your field. So if you look at the development over time, this was a snapshot in 2015, as you can see. If we look at the development over time from 2000 to 2015, the graph is a little bit confusing. Maybe we have two axes here, one on the left-hand side and one on the right-hand side. So the first three indicators here, drought, malaria and the broad definition of conflict, we see a slight increase in, oh sorry, this is drought, we see a slight increase in drought exposure during this period but all in all it's relatively stable. This is, I should say that for drought and flood and the broad conflict definition here, we don't allow the hazard to change over time, we create this hazard map and then we look at how the population changes affect the way the shares of people are exposed. So actually the main thing here is probably that the number of people, although these percentages are smaller, the number of people exposed to floods or the share of people exposed to floods is increasing. Share of people exposed to conflict is also increasing a lot whereas the narrow definition of conflict, this risk of being killed first drops from 2000 to 2005 then increases later. So in this case we allow the hazard to change over time in this conflict B scenario. And also with the malaria we allow the hazard to change over time since this has been something that has changed a lot. So essentially the climate variables we don't allow to change over time because it changes more gradually whereas the human made risks such as conflict and malaria we allow those to change over time. We see this dramatic drop in malaria incidence from 2000 to 2015. Now these are the shares of population. What happens then if we look at the number of people exposed? This has been indexed here to 2000 is equal to 100 and then we look at the development over time. So we see for most of these, for these three that are only dependent on population movements and the fixed hazard, we see that droughts, floods and the broad definition of conflict there are around 50% more people exposed to these hazards now than in 2000. Malaria if you look at the number of people exposed to a prevalence of 10% is relatively stable over the period. And again the number of people exposed to conflict goes up and down because the conflicts end and start in new places and stuff like this. Finally I wanted to show you this table which is a simple correlation matrix between the different types of hazards and exposure to these. And I've added this column with whether or not a pixel is urban or not. So we can start to draw a little bit conclusions about what is driving these changes in exposure to hazards. We see that there is a negative correlation between being in an urban pixel and being exposed to drought. This probably makes sense because there's less drought close to coastlines, close to rivers, irrigated areas. There's a positive correlation between being in an urban environment and being exposed to floods. Again urban areas are located typically close to rivers or coasts and these are the places that you might experience a flood. Also this is the broad definition of conflict here in this table which also increases dramatically if you are in an urban area compared to a rural area. This is because the chance that there has been one conflict event in your city if it's a big city with at least 10 killed is rather big. So if you just live in this city you have a bigger risk of being close to a conflict. This does not mean that you have experienced a bigger risk of being killed in a conflict on the other hand which doesn't have the same correlation with the urban rural variable. There's a lower risk of being exposed to malaria if you live in a city than if you live in rural areas. This is also in accordance with the medical literature simply because these mosquitoes that carry malaria have worse living conditions in places that are developed. People live indoors. If you live indoors in concrete buildings then your risk of getting malaria is lower. This is the way it looks and so the conclusions are that if you look at the share of population exposed to these risk factors the share of people exposed to drought is relatively stable over this 15 year period whereas it is increasing for floods dramatically reduced from malaria if you look at the risk of being exposed to conflict it is either very volatile if you allow it to change over time or increasing if you just look at whether you are close to a place that has experienced a conflict during the period. The number of people for almost all of these risk factors is increasing a lot except perhaps the exposure to malaria which has been stable in this case and probably the main driving force for this development in what we see in this paper is urbanization. People moving to cities become less exposed to drought, more exposed to flood, less exposed to malaria, more exposed to conflict and of course a future research topic would be to cover this with the potential impacts of climate change to see how the numbers increase if you allow temperatures, precipitation or other variables to vary over time. So this was it. Thank you.