That work done. All right. Yeah. Go ahead. Yeah. No way. Though. I want to introduce the diversity minutes and I tried to find something that's actually going to lead into the talk though. On April 7th, the EPA announced that the agency is going to take actions to advanced environmental justice. Though there was the Biden administration asked for all agencies to start considering inequality in everything they do. And so the EPA, It's going to tackle environmental justice and they pop a couple of steps they're going to take, including meeting with marginalized communities. And as well as I wanted to mention, both the Senate and the House has created committees on environmental justice and the co-chairs of the Senate committee was our very own Harper and neighboring Senator Cory Booker. And I thought that would be interesting folks. But it leads into our talk today. Because part of tackling environmental justice means being able to figure out when and where it's happening. And in some of the mechanism though, they, we have on our overall umbrella. Who's a postdoc at New York University, to be an assistant professor at the University of Amsterdam. And she's going to present my rule of thirds of inequality. How different neighborhoods flooding though? It's I think a knife appropriately done. Okay. Thanks very much, Kimberly. And thanks again for inviting me. I'm very excited to how the opportunity to share my research with you. So as Kimberly mentioned, I'm going to present a project title, search of inequality, how different neighborhoods react to Florida. And what this paper does. It investigates how different places with different characteristics and different types of residents react or behave after flooding. And what I'm going to show you is that these responses were very, very heterogeneous. And this heterogeneity lead to poor places becoming poorer, enrich places becoming richer. So in a way this is flawed. These natural disaster acted as a searcher waiver increase in inequality. So let me start by motivating why exploring this question, this particular project is interesting and relevant. Floods are a costly natural disaster. Some estimates by the United States Geological Survey estimates that flooding was the most costly natural disaster during the 20th century in terms of lives lost and property damage. These are, these damages are expected to get worse as we move into the 21st century. And this is specifically in the North East US you to sea level rise and changes in Curie connectivity. As a result, there has been a large literature, specifically in urban and environmental economics, understanding how, how real estate, how neighborhoods behave after flooding. And they have found, this evidence has found at different results in different contexts. So for instance, or data and tax benign who explore property, real estate property prices after Sandy Hurricane Sandy just in New York City, find that property, Florida property, so a decrease, a permanent decrease in prices in a different context. Amino Landry in North Carolina find that prices are flooded, properties go down, but they revert to previous pre-flight prices after a few years. And yet in a different context, yarn Panofsky, who studied the effect of floods in the state of Florida, find that prices go up after a flood. To see differential responses to fluorine, we can even go to a more micro neighborhood scale. In here, I'm showing you, I'm highlighting you to counties of the New York City area that were affected by Hurricane Sandy. On your left, you see highlighted the borrowed the county of queens were flooded properties. So a decrease in prices of 19% compared to nearby non flooded properties. On the other hand, the nearby Hudson County in New Jersey, similar distance to meet a Manhattan similar, similarly affected by the flood, the flooded properties. So an increase in prices compared to nearby not slaughter properties. The reasons why some neighborhoods react differently or behave differently after the Florida have been under explored in the economics literature. And moreover, is still not well understood how this effect compound on existing spatial patterns of gentrification of urban decay or a rather countercyclical to them. This is at a time in which many disciplines that economics and in particular is exploring the negative effects of spatial segregation. Not only work by Raj Chetty and coauthors that find that the place at a very micro neighborhood level the children grow in have an effect on, on outcomes in adult life. Specifically, they find that children who are raised in neighborhoods that have less concentrated poverty unless income inequality yield better outcomes like higher wages when they become adults. There's also growing literature that investigates the feedback loops between spatial segregation and economic inequality. So in this context, this project is going to ask whether the changes in real estate prices in a neighborhood composition that the flawed cost has contributed or rather counterbalanced these existing patterns of a spatial polarization. For the setting I'm going to obscure this questioning is Hurricane Sandy which affected the Northeast Coast of the US at the end of 2012 into date, remains the fourth most costly natural disaster in US history. After Hurricane Katrina, Harvey and Maria. The hurricane Sandy's a compelling starting to explore heterogeneity response as a response to flooding because it affected a wide Band of a geographical area in the, in the, in the US. And to explore these questions, I'm going to compile a breach dataset that includes the universe of property sales, a subset of which I'm able to combine with race and the income of the buyer of the property and a comprehensive set of of characteristics of the places that were flooded coming from the census and other, and other sources. So just as a preview of the results that i'm, I'm going to show you what I find is that indeed there was a heterogeneous response to Sandy's flooding. And specifically, prices and prices went places that were high income to start with have higher property prices and a large share of high income residents. So an increase in property prices after the flood, whereas, whereas flooded properties in low-income places, so a decrease. So these messages summarized in this, in these two figures, I'm showing you both figures on the one, on the x axis, plot the time, and then on the y-axis, I'm showing you the evolution of, of, of property prices after removing block fixed effects are basically variation around the census block mean. And in both graphs, the blue line shows the evolution of prices of properties that eventually we're inside this floodplains obesity affected by the flooding. And in orange where nearby coastal properties which will not affect. The graph of your left, is summarizing this price evolution for the poorest places in my sample. Basically the bottom decile according to their income or the income, the average income of the plates. And on your right is the evolution of prices for the high places. And what we can see is that in the poorest of places before Sunday, both flooded and flooded properties followed a similar trajectory more clearly. So after the financial recession in 2008, then when Sandy heat at the end of 2012, the prices, the evolution of crisis clearly diverges and properties that were flooded. So a drop in prices that remains below those of non flooded properties for the four years for which I have data. The behavior of properties in high-income places is remarkably different. Before the flood, we see that the, the, again, the property prices of flooded and flooded areas follow a similar trend. But then when Sandy hit days, no appreciable, no drop in prices for flooded properties. Not only that, but the prices of a flooded properties are both those of non flooded for the four years after, after somebody something that hasn't happened in the 10 year prior. To the remaining of the talk is going to follow the usual outline. Let's start by motivating why it would be rational to respect and increasing property prices in some places, specifically those that were high-income to start with. Then I'm going to describe the data, the data sets I use in this project and the models. I, I employ to analyze the data. Finally, I'll present, at present the results of this analysis and alcohol. Okay, so let me motivate, present a framework of thinking of why we would expect property prices are flooded, properties to go up after the flood. To that aisle, I rely on the canonical model of segregation developed by Gary, Becker and Murphy. So this model assumes that a household, the willingness to pay for household to live in a specific, please depends on two things. First, the, the exogenous are many disciplines. Has say a nice beach in a, in a, in a, in a coastal location. And second, the type of residents in, in that neighborhood is specifically, it assumes that are positive externalities derive of living with high-income people or other characteristics that are associated with income such as race. These externalities could be a pristine signal, networking opportunities, et cetera. So under this model, in equilibrium, we should expect property prices to be higher in places that have high, high nice humanities and, or a high number of high-income vessels. I then I assume what our See what happens to this system in equilibrium when IS shock by afford on their two assumptions, I assume the flood is a decent money for everyone. No one likes their house to be flooding. Than second, I assume the flawed as a, as a, or I conceptualize the flood us a wealth shock and assume a decreasing marginal utility of wealth. What does this mean in this context? It means that low income residents, the right, the higher this utility from flooding, because they're less able to reconstruct after the flood, ness able to retrofit their properties after, to be prepared for the next flooding or unless able to ensure proper defense funding. Moreover, I assume this this implies that the dispute for a given type of precedent, that this utility given by the flood is higher, the more expensive your house was before the flood. Again, thinking that more expensive houses are more expensive to ensure. So with this context and this assumptions the model yields to make predictions. First, that it would be rational to, to see low-income residents opting out from leaving him. So floated places, specifically those that had high preexisting or higher or property prices to start with. Hi income residents dislike their neighborhood a little bit more than before the flood because they don't enjoy the flown in the flood, but they're better able to stay to coop and staple. As a result, in this neighborhood, we see a higher percentage of high-income precedence. And then there are two forces at play. On the one hand, this neighborhood has been flooded and no one likes the flood, is this eddies are many. But on the other hand, there's higher percentage of high-income precedents and the model assumes everyone enjoy, so they're positive externalities derived of living among them. If the second force happens to be higher, it will be rational to see. Property price increases after the flood. Even if we assume that everyone is rational and everyone has accepted or internalize the flooding knows the next slot is going to come. So after, after this, again, I'm going to describe the dataset I used to test whether the predictions of this, of this model actually hold in an empirical setting. And let me also say if you have any questions while I speak, please feel free to jump in and do so. I I I don't think I can talk and look at the chat at the same time, but if you jump in, I'll I'll I'll answer the question. Otherwise, I'll leave time at the end for questions. Okay? So the, I use through the main data's, I'm going to describe the main datasets I use. First. I focus on the four states more impacted by Sunday, those where New Jersey, New York, Connecticut, and right. So for these four states, I've, my dependent variable of interest are going to be property prices and end. And at the socioeconomic characteristics of the bias of this properties. This comes from. So these dependent variables comes from several data sets. The property sales come from a proprietary vendor. Core logic. Basically, what core logic does goes to all county clerk's offices and collects these data that are stored. There. Basically property, real estate transactions, which are publicly available but sometimes seeding, you know, in obscure PDFs or even papers. What they do is they collect this data, they digitize it and then sell it at a hefty price MSA. So the, the so they collect the universe of real estate transactions. What I'm going to explain in a minute for identification, to identify the causal effects of the flawed. I'm focusing on and on a narrow band along the coast. Easily compare properties that were flooded with nearby no flooded properties are marking the location of my of of the of the property sales for which I have records as orange dots in this map and you can see how it covers the whole coast that was affected by by Sunday. I'm able to merge a subset of this data of the property, the property sales data, with the raised an income of the buyer. And this is using data that is publicly made available under the Home Mortgage Disclosure Act and using a procedure that has been used before in, in the economics literature. Then I also compile a rich set of characteristics that describe the places that were flooded, the income of the place. And another country such as taxes, the segregation of the place, transport, connectivity, et cetera. This data, the choice of these variables comes from exploring the urban economics, the public finance literature determining what are the characteristics that determine whether a household, the sites, where to live. Then. My main explanatory while I have one quick essentially revise it again. Yeah, go ahead on the eye. There are a fair number of people, especially in coastal areas that will buy properties without mortgages. Especially higher n, yes. So it seems like those are kind of censored in this dataset. Indeed, during emerging. Yes, indeed. So the, the residential property sales, basically the core logic dataset, contains all of these transactions. Everything is is is recorded. But you're right that the Home Mortgage Disclosure Act only includes data for residents who ask for a mortgage. So what I do in this table here, I present the summary characteristics along those variable of interest rate, the sale price, the square feet of the properties for the full sample. And therefore the subset, the same characteristics for the subset of the, of the properties that I was able to merge with the with the race and income of the buyer. I'm able to merge 30 percent of the total of the total of the total number of observations, a half, which is approximately 50 percent of, of, of those are sebaceous to have mortgage data in the core logic dataset. This is similar to what other researchers were able to achieving a different, in a different context. So what we see is that indeed, following your intuition, the sale price of the full sample of properties is slightly above those of the of those that I was able to merge with the mortgage data which would be indicative of of of high properties with higher price being sold without a mortgage. So indeed, this is a caveat that needs to be taken into account. And as I'm as I'm going to show us, see that in some places, high-income places, we see more high income white buyers buying properties. And I would argue this is probably a lower bound of the actual change in demographics because a lot of this high income, presumably white buyers are not recorded as having bought a mortgage. But i e, That is the point and definitely a caveat of this, of this, of this data. So i then I can thank you. Thanks for the question. I'll, I'll I my explanatory variable of interest is then the, the flood extent. And for that I use I use maps of the flooding extent and depth that were developed by fema. Fema constructed these maps using several data inputs like satellite images, a field B sets and Bieber and rubber gauges. And I'm showing you here the map and as you can see, the area affected by the flooded for the, for the flooding when from Rhode Island to New Jersey. And and I have a very detailed map of of, of which properties where affected and, and by how much. So these are my main data sets and now let me quickly describe how AI models are used to analyze them. First, as I, as I anticipated, I'm going to be focusing on properties that are near to the coast. There is a wide literature exploring how people who self-select to live on the cause may be different from people who choose to live more inland. Say for instance, on their degree of risk aversion or they're believe that climate change, sea level rise is actually going to happen. So to make sure I compare apples with apples, focus on properties that are located near the coast. So specifically, I construct a buffer from the from the shoreline as the fight by noaa, a buffer of 500 meters. So 500 units is a distance one can walk in 67 minutes. And what I'm gonna do within this buffer along the coast, I'm going to compare properties that were flooded. So in, in here in this map I'm showing you with different levels of green and yellow. The properties that were flooded and the properties are the orange dots. So I'm going to compare these properties that were flooded with nearby properties located no more than a 100 meters from the coast, but we're not flooded. So then the first step I'm going to do is to evaluate average average impacts of the flood, mostly to act as a benchmark of the heterogeneous effects I, I measure later. In for that I'm going to use a difference in difference model. So basically I'm going to compare properties in the floodplain and outside the floodplain before and after the flat, you're going to see how the flood affected affected the group of properties that were flooded after the flood. I'm then going to be interested in measuring how properties located in different places with different characteristics had different effects. So for that, I'm going to use a triple difference model, which is basically comparing properties before and after the flood in the floodplain, outside the floodplain and with a different level of a specific characteristic, say the income of the place or the education of the, of the place, et cetera. The third and final method I'm going to use to analyze this, this data. It's rather newly developed machine learning procedure. This, this type of procedure, I'm going to specifically one developed by Victor. She's not Zhukov. Start to flow in coauthors are becoming more minutes, more and more used in economics. Others used for it's developed by Suzanne assay at Stanford. The, the, the, the value of this, of this procedure is that it would allow me to explore whether there is heterogeneity in my results without making any assumptions or where this heterogeneity is coming from. Just to compare with my previous method in, in, in, in, in, in the triple difference model, I'm assuming that stay the income of the place, as is determiner is driving different results with this procedure, I'm going to remain agnostic and I'm going to let the data tell me what are the variables of the place that better discriminate the places that were most and least affected by the flood. In very succinct terms with this procedure does, is for every observation in my sample, it creates a counterfactual for itself. Basically what would have happened if this flooded property was not flooded. Then I can group the properties are in my sample according to this predicted treatment effect and determine which were the most affected and the least affected by the flood. If the results were homogeneous, I should expect this effect on the most affected, enlist affected to be pretty similar. If they are not, I can conclude days heterogeneity. Okay. So again, just to act as a check of where the heterogeneity coming from without making any assumptions about it. So without that, let me jump into present the results. I'm going to first present the results of the of the average effects of the flood, then the results of the machine-learning algorithm, and then finally, the results from the triple different spiral. So I have another quick question. Sure. Yeah. So it's not repeat sale data, right? You're not you're not doing the same property sold before and after in the treatment area and control areas. It's all the properties and you're using fixed effects to control for yes. So that's yeah, that's, that's a very good question. I, I, in my main model, I'm using fixed effects, census block fixed effect. So basically I'm comparing them. Property sold within the census, the same census block through time. Now you have enough sales to try repeat so, so I, I do, and I check that the obesity and repeat sales model, you will be comparing the same property before and after and NLST that the results are consistent and robust to including instead property fixed effects. The reason why is not my main model. It's because I'm exploring a 15 year period from 2002 to 2017 and ends for repeat sale to appeal, my dataset would have have to be sold at least twice within this 15 year period. And I worried that property that has been sold so repeatedly in in this in this relatively short period of time, may have some unobservable characteristics may die in the results. So I, I, I included as a robustness check, but not as main as the main model. Now, you may be worried that properties that are sold within a census block, even if it's a small area, might differ in say, properties that are sold at some before and after the flood are different in the sense of the size of this block is pretty small. Here is, please. Yeah. Exactly. So that is that is that alleviate some concerns? The fact that it's a small area. But just to alleviate concerns even more, I in some other models, I also include individual characteristics of the property, say the year it was spilled, the footage. And and just to rule out that it's it's the the different characteristics of property salts within a census block that is driving the results. Thanks. Okay, so jumping into the results, I, I, first, I'm going to show you average effects of the floods in here. I'm still not exploring heterogeneity. This graph is, is, is plotting the same data show you at the beginning, but for all, all properties in my sample. And the y-axis again is the, is the sale price after removing block fixed effects in logs. And then on the x-axis is the time and the dotted line marked when Sandy sandy hit. And blue line is the properties that were inside the the the flood plain, eventually inside the floodplain. And while we can see, is that on average, again, the property that way in the floodplain right after Sandy. So a sizable decrease in prices that remain below those of non flooded properties for the four years after, after Sandy hit. I then explore the flawed a second teen years as a continuous measure. How I do that. I I oops. Let me back. Sorry. This is okay. My my screen froze but it can you see a little map on your left and a little weekend? You can sing. Thanks for confirming. Yeah, my screen froze. So what I do again now it's exploring the flood, but looking at it as a continuous treatment in the, in the previous graph, I will show you, I always assume in the flood us that dichotomous variable, you either were flooded or you were not. Here. I'm allowing the floor to have a different effect depending by how much your property was flooded or by how much your property the flood. So to do that, I'm here on your left. I'm zooming in on a on a on a specific area within my area of study. And I'm plotting in blue the maximum extend the flood reached. This comes from the femur, the femur data and then the orange dots are residential property sales in my sample. So what I do is for each property both flooded and flooded, I compute the difference in elevation with respect to the closest point in the flatline. So what this allows me to do is to, is to generate a variable, a continuous variable that measures if your property was flooded, by how much it was flooded, and if your property escaped the flood, by how much it escaped the fire. Okay, So then I'll, I'll allow some again, the flood to behave differently according to the, extend to the depths of flood or escaping the flock. So on the, in this graph, on the x-axis, I'm plotting this distance to the flood, that this vertical distance to the flood that I just explain how I how I did. Okay. So the dotted line is the maximum flood level that the, the reached nearby, nearby properties. Then the positive values are the showing the depth or the extent to which non flooded properties escape the flood. And negative values are the, the depth of the flood for those properties that were flight. On the, on the y-axis, I'm showing again the sale that the log of property sales, but I'm also removing hear variation that comes from year to year. So I'm not planning time. In this graph, the orange line is uploading property prices before Sandy took place. And are in blue are the evolution of prices for properties after Sandy. So this graph shows that for properties that were not affected by Sunday, even if glues two to the two Sundays flood, we see no difference before and after, right? So the, the, both lines follow very similar, very similar trends. And we can test that there is indeed no statistical difference on how these transferable, on the other hand, flooded properties after Sandy's t. A remarkable change in how, in how prices above. We see that even for properties that are only mildly affected by flooding less than half a meter of depth, we see a decrease in, in property prices. We see that evolution in prices follows somewhat U-shape. This could be due to coastal properties that were more flooded, had a higher flood depths. Are those located on the sea front with ocean view thing we can could be argued that this is, there is still a premium to the sort of properties. So then I then move on to explore with this machine-learning procedure whether these average results that I just showed you before, that, sorry, let me, let me, until now I just show you the effects on prices. And, and this graph over here is, is as putting a number on that. I see that the fluid properties after Sandy experience a 9% decrease in price as compared to nearby no flooded properties. I'm also now showing you how the race and income of the buyer, so this property changed. And what I find is that these buyers are less likely to be high income and white. This is a composite variable. Composed by a principal component analysis that basically ranks, or if you rank high on this variable, means you're more likely to be a high income white buyer. It has been normalized to have a mean of 0 and a standard deviation of one. So basically the message is that on average, flow properties have lower are sold at a loss with lower property prices and the buyers of these properties are less likely to be high income and white. I then move on to explore if this average results are masking heterogeneity using the machine learning procedure I briefly described earlier. So what this graph is showing you is the average effect of the flood on properties on the five groups. Properties in my sample, ranking them according to their predicted treatment effect. So the, the, these machine-learning proceeded also yields the average treatment effect, which again is similar to what we find, what I find before. Basically that on average property flooded property prices, so a decrease in prices of 9%. This gives me confidence that this machinery procedure is measuring what it should. But then it also shows that be this average is masking important heterogeneity. The group of properties that 20 percent of the properties that were the most affected by the flood. So a decrease of 30% in prices. So significantly higher damage, if you like, higher shock. For these properties. On the other hand, there is a 20 percent, the 20 percent least affected by the flood. That so an increase in prices. So these properties is 20 percent of property. So a 20 percent increase in prices compared to nearby no flow properties. So the flood clearly created winners and losers even if the average shows an average decrease in prices. This procedure also allows me to go ahead and explore them. What are the characteristics of the places these properties that were the most affected and the least affected located in. And what are the characteristics that better let me discriminate the, the sort of places differential effects took place. And this is summarized in this slide. So in each row i'm, I'm denoting characteristics of the, of a place. I am combining different indicators for each of the category. And this is following the work of cherry and others with a principal component analysis to basically get a variable that summarizes different indicators, tells us a story on the second dedication of a place, the income of a place, the retail availability. All of these variables have been normalized to have a mean of 0 and a standard deviation of one. Have a rich data set of 12 of these characteristics. What I'm showing you in here are the five, please characteristics that have the largest differences, average differences in this for the properties that were the most affected by the flood, that lost the most after the flood. And those properties that we saw that again after the flood. And what this analysis does is it finds out that that second, that income, the income of the place and secondary education, which has two variables that are very correlated or a high correlation in my, in my data. These two variables yield the largest differences in-between these places that were the most affected and the least affected. In this context in my data set, other characteristics that having posited as having an influence on how fast places recovered after the flood, like different measures of social capital, do not appear to be as good as discriminating which places are the most or least affected as income or secondary education. So this would be consistent with the model or the frame where I just told you earlier of high-income places not only been less affected by the flood, but even at some point, gaining after the flood? Yes. Or in back and explaining the segregation and migration variables. Yes. So the, the these two let me see if I can. So in the eye, again, all of these are composites of, of different, of different indicators. So in the, in the, in the segregation, in the segregation variable, I'm including the racial shares of a census tract. So basically share of white, black, Asian, and Hispanic in, in a, in a, in a place then the share of, of commuters with less than 50 minute. This has been determined in the literature as being indicative of the simulation of a place. And then I have other measures of racial segregation, income segregation made available by the opportunity grouping in, at Harvard University. So basically these are the variables that are included in this, in desegregation. Composite. And, and then the migration variable that you also mentioned. Ai in this variable measures the inflow rate of, of, of, of basically residents coming into the county doesn't have to be foreign residents. It's just people moving into this county and then people moving out and then Isles. It also takes into account or, or, or, or summarize the fraction of, of, of foreign born, basically non-US residents living in a census track. Thank you. Great. Thanks. Okay. So Okay, so again, we have determined that this income and secondary education or the variable that better let us predict where we expect property price gains or losses after the flood. So then I go ahead and take this income variable to, to explore in a triple difference analysis the different evolution in prices and the socioeconomic factors to solve the buyers for properties in, in places with different incomes. So this is the graph I show you at the beginning, which is again plotting the evolution of property prices over time. Properties, None a non flooded in orange and flooded in blue. And on the left is the poorest, the properties in the poorest of places, the bottom 10 percent, and then the richest of places. And I say essay, as I mentioned at the beginning of the, of the talk, we see a clear large decreasing property prices for soil properties in the lowest income places. But that shock is non-existent, basically non evident in the high-income places. Not only that, but we see that the prices of our properties are above those of non flood properties. I then also explore changes in these two types of places with the continuous, continuous measurement of the flood. Again, on your left is our properties in the poorest of places. And on your right on the richest of places. Can see, in both cases, properties before and after Sandy that were non flooded follow pretty similar paths. We can test this statistically. They, they are, they are indeed not following different paths. And in the poorest of places we see how flooded properties. So a decreased compared to non flooded properties in this poorest of places. We don't see this uptick that we saw on average for four more flooded properties, basic, arguably properties located closer to the sea, to the ocean or with ocean view CPUs. On the other hand, in, in, in high, in high-income places, the results are noise here, but we can see that specifically the four properties that were not affected much by the flood, basically with lower flood depth flow properties see an increase in prices compared to non flooded properties. I then move on to explore the changes in in the type of buyers of these product. On the, the, the, to the the top panel. The first of this figure is reminding you of the average effects for the whole sample that I showed you earlier. Is the, is the turnover, low income, higher income area about the same, or are there the share of sales larger and one versus the other, I guess. And that's indeed, that's a very good point I do I do evaluate the likelihood of of of of sales. Basically, how likely is a house to be sold in low-income places, high-income places before and after the flood. And I do not find. Differences in, in, in between both. I have some, I have some graphs that I can show you in a minute about this. So I'm kind of wondering if there's almost some, some sorting that's going on. Maya's way in some of the like. Like everybody sees these as riskier areas after Sandy. But the risk preferences of high-income people in low income people, maybe they're really different like high income people are more likely to take on take on risks. Yes. Who are in SI? Yes. Explain some of the spread maybe. Yes indeed. And I do think that the, that the 14 story that, you know, it's evidenced by the fact that they different types of people are buying these properties is as, as behind a big part, a big part of the story. We see that high income residents are, are more able to stay in an area because they can, uh, for reconstruction by flood insurance, whereas low-income residents are or not. And then this, this, this high-income places as a result of the sorting become endogenously more attractive? A higher percentage of I can't write capacitance. Anna. Yeah. Yeah. You you mentioned a number of times that the flood affected, but really all we know is that there's a correlation and it may be something in the high-income properties. If my property say my kitchen was flooded, I may then spend $7,000 thousand to remodel my kitchen. And you don't capture that. Says I understand it. Whereas my neighbor property wasn't flooded. And so he or she doesn't invest the seven the 70 to $100 thousand to do this remodel. And it may not have anything to do with the flood per se other than I mean, the reason these properties maybe more. They're selling for greater amounts. Again, because people have invested and we put money into these properties. Yeah. So that that's that's that's indeed a concern. I I do. What I want I can test as which is partially answered. You're addressing your concern. Female, besides the flatMap, has has categorized properties according to their damage. So they have Useem Useem Useem, objective physical estimates of the damage and they have they have declare a property has been effect that having minor damage, major damage or destroy. Looking at say, if 20 percent of your roof miss him, but you're not missing any walls than your houses consider having major damage. So what I can what I can do is I restrict the analysis to properties that have the similar level of damage and basically seen whether properties that were affected, say declared as affected, whether this property is half. Whether the income of a place is still determining some heterogeneity. And I find that they do. So basically, it's not like properties in high-income places where less damaged, say, and that is driving the results together. Theorem is question 1. What damage though? Because homeowners now you put money into kitchens and baths and you can get it out. You put money in there, everything else. And you mostly just do it for yourself. I mean, you're you're Dan is damaged and maybe you've gotta fix it up some. You're not necessarily going to be able to recoup that in a sale. So, you know what's damage matters? A lot. Yeah. Yeah. And and absolutely. And so I I I'm skeptical without younger ending, what people did. Whether this, what the effect is. It seems like you could put in a variable distance from the coast. So yeah, yeah, a difference. In the IEEE. I also have results with horizontal distance to the flatline. And, and, and find similar results than the vertical distance. I I thought that the vertical distance was in terms of the depth of for what property was started or escape the flood was more what I find is more correlated with damaged. I could also easily do the distance to the coast that you're suggesting that I haven't I haven't done yet, but that's a very good suggestion. You think improvement's data like, yeah, you might like the record building in a record. Exactly. And that's that's yeah, that's available. And indeed, and indeed, I mean, you touch upon something that has come up for people who are give me advice on this project indeed, a few times. So that is definitely the next step on, on this project. Did there there is detailed data on the properties that ask for reconstruction permit, so I could be able to see whether there is a difference in that. And and and dad say if it's reconstruction that is driving the results, it would be consistent with the story that this reconstruction is only taking place in some places. So it's still consistent with the fact that we are seeing this divergence them basically this reinforcement of existing equilibrium by the flopped. Some places, high-income places are better able to flood came and now I have moles, I'm going to change the carpet or, or redo the bathroom. And now my properties more expensive and more attractive than n. Whereas the flood is leading to some places to go more into the K. But indeed I, I, it's a point well taken that I will need to provide empirical evidence of fat. Okay. So I'm on the eye, you have 10 minutes left. So let me also tell you some other analysis that that I've done and some have come up in questions and then maybe if you have specific questions about those, we can we can target those. You may be also wondering what's happening with flood insurance. Flood insurance definitely has a role to play when we think about reconstruction of, of, of, of a place. So I, I, I explore changes in inflow insurance using the universe of of, of of policies and claims under the the the National Flood Insurance Program that fema made publicly available a couple of years ago. So what I'm showing you in this graph is the evolution on the number of policies that in, in log after moving sensors, truck fixed effects. And in orange I'm showing you the evolution on the number of policies for the low-income places. Basically the places they saw property price decline and a lower and lower number of high income white buyers. And then in blue, the evolution on the number of policies for high-income places. The first dotted line is when stand the heat at the end of 2012. The second line is one year later when many policies, many flood insurance policies would have to be renewed if they were signed like one year right after Sandy. And then the third line is three years after Sandy, which again is a common renewal period for flood insurance policy, where we can see both in low-income places, in high-income places is that right after Sandy we see a search, an increase in the number of flood insurance policies. There are science, this has been found in the, in the, in the literature before. This is, this is a common response to fly. But then one year after or soon after one year when policies have to be renewed day or not. And we see a decline in the number of policies both in high-income places, low-income places. However, this decreases, this drop in the number of, of policy insurance number of policies of flood insurance in, in low-income places. This drop is so large that I, I it results in fewer properties having flood insurance three years after Sandy done before sunk. On, on the other hand, in high-income places, they can not rule out that the number of policies is different three years after and before and before. So this is again basically communicating the same story. This is low-income places are not only being sold to lower income individuals and have lower property prices, but they're also less likely to be short after the flood. So we show in a vulnerability in, again appearing or emanated in, in those places. You may be also worried or thinking about that. Maybe this high-income places were also able to attract more funding for reconstruction and that could be leading. Some of the results I find I using data from the public assistance program at female, which is the largest pocket of money for reconstruction after a disaster. I do not find that high income places are having more money per capita. Now let me also say that this data is at the county level. So I'm making inference only based on 33 observations. So we should take these results with a grain of salt. But they seem to suggest that they are not, there is displaces or not indeed receiving more money for reconstruction on a per capita basis. We talked a little bit about this when thinking about the type of properties that are being sold and is that high-income places are more able to reconstruct. I do several things to explore this, I find that there is no more new housing. Basically a sort of creative destruction sort of story. The flood queen, a tabula rasa in which to build better, more attractive housing that is attracting high-income individuals. I do not find evidence of that. We also and this is and this is going back to George question or are there more or fewer houses on the market after the flood? Graphs are showing you the likelihood of the basically the percentage of property sold in a, in a given month, the evolution of of that probability in blue for priorities that were eventually the floodplain in an orange outside the floodplain. And then on the on the on the graph on your, on the top right is for low-income, low-income neighborhoods. And then on the bottom high-income neighborhood's. And we can see there is a slightly upper upward trend, meaning that properties on that were flooded are more likely to be sold, not an average but on, on the trend is positive. But this positive trend is happening both in low-income places and in high-income places. So it doesn't seem that a differential trend is as it's driving this, this result. I again, if you have a question, we can we can we can discuss I also find that the results are not driven by property that had recently been flooded a year before by Hurricane Irene that affected New Jersey is not like these properties where aware of the flood risk and they behave differently. I also see that the properties that were labeled by females, high risk, even if they weren't flooded also saw a property price increase. But the fact of being on a female risk area and flooded by Sandy compounded. So basically these are two different effects that compounded and, and and, and contributed to the changes in prices. Do not find that there are more renters in place after flood. I also do not find that there are more primary residence in low-income place. Some other worry you might have is that This high-income places are mostly beach houses that high income residents go every once in a while. So they are not really aware of the Florida or they do not internalize it as much. But actually find that in high-income places are more properties that are primary residences. I then do robustness checks to try different alternative specifications. I also see that the property see my control group. Basically properties that were non flooded by a nearby not see changes in prices or in the type of, of, of by your thereby in these properties. These rules out that our changes are spillovers into the Properties under control that are driving the results. Okay. So with that, let me conclude. So what I just show you the result, I just show you today evidence that the flooding caused by Sandy deepen or exacerbated existing spatial equilibrium. So high-income places were less affected by the flood. Where so an increase in prices, more likelihood of being bought by high, high income white residents were low-income places follow the opposite trends. So there's, this original source improves our understanding on these differential trends we saw after places were flooded, dependent on the, on the underlying characteristics of the places that are flooded, we should expect different behaviors. So these, these results have important policy implications. So when a place flooded and what these results evidence is that there's a duality at place. Say there is. On the one hand, we see some encroachment of some places of the vulnerability of some places, places that were low income to start with see a decrease in property prices. Buyers who have less lower incomes, arguably than less able or less resilient, the face of the inevitable next Storm. Have nowhere or OPT less to have flood insurance. On the other hand, we have at the same time reaching claims, becoming richer. Now, one could argue this is just the market operate in if they want to live on the coast and if they take the risk, then so be it. But then we, the political economy of the situation is that they can displaces, can then, when the next storm comes, when the next Hurry can come, they can use the higher political clout to regressively soak up. After flood recovery. There's it comes off and on the news, reach communities in Florida arguing or lobbying for their roads that go to the properties to be raised that cost millions and millions of tax payer dollars and, and, and, and, and one could argue that this money could be spent elsewhere. So basically. Is this duality at play? What is the policy recommendation that will depend on what our policy objective or our goal is. If we are worried as we should be, that sea level rise is happening and that people should retreat from living on the coast. Then we should think about incentives for this high income residents to opt out from living in this, in this risky areas. On the other hand, if we are worried about existing current inequality, then we should look at this low-income places and, and, and think about how to make them more, more resilient and how to eliminate the, the, the pockets of vulnerability that, that the flood generates. So that was all I have prepared for today. Again, thank you so much for your attention. I went two minutes over time. I apologize for that. If there are any he still any questions I can take them now or I'll be glad to chat afterwards. I can maybe. Now I see one in the chat window on housing prices go up and flooded areas becomes more costly to bio. Those homeowners, which has big implications for government official. Bethany. So the buyouts, yeah, the buyouts is definitely an important part of the story. In the US, I have a large dataset of property sales. I have, as I show you, 0.5 million observations, the the the buyouts that happened after Sandy where small percentage of those. So I I didn't single them out in this, in this project. But indeed it has. And there's a large literature exploring that in it's indeed an important, an important aspect to think about whether you had displaces high-income places becoming more soluble, again are highly, are becoming harder and harder than buyout. And, and if, if a if a retreat is where we're heading towards, I can jump in. I was wondering or thinking about massively places like Houston, for example. A lot of the flooding wasn't necessarily it was like move or geographic specific eye relative that bill and unnamed woman there, one of the few houses in the neighborhood that I was wondering what your expectations might be just because of the the things we were mentioning in terms of distance to coast and sort of the desirability of an area I went through, Houston could be a good sort of comparison in terms of what happens in general for income and expense flooding. A lot of those have to be completely renovated. But from what I what I earn just anecdotally, a lot of people actually didn't move because a lot of those neighborhoods were actually written up. Yeah. And indeed that's, that's, that's a very interesting context in the sense that the amenities of the flooded places are less good. In the sense there is Nora and non-core such as ocean. But there is evidence again coming from Houston as, as, as you suggested of, of, of, of these dynamics taking place as well of flood happening. And, but high income residents wanted to, first of all, high in compression is less likely to be flooded, but if they wear, more likely to go back and rebuilt and only them, so then that neighborhood becomes more higher income must have results and arguably more attractive for other higher income to the income and leaf there. So there is, again, I believe the results are extrapolate. Again, I haven't tested them, but based on an antibiotic all evidence, this is the case of flooded places becoming more soluble where others, low-income places becoming more abundant, more foreclosures, people less likely to move in to moving. They're talking about other contexts. There's also evidence in California with the wildfires. After wildfire, high income residents more willing to go back to that area and rebuild dan low-income residents so it doesn't even have to be flooding. Other natural disasters that affect residents differently may have a similar effect. Paper over I meant over. Thank you for staying. And honor, thank you so much for presenting this very detailed, well-thought-out paper that influence no much data. And just thought.
Ana Varela Varela Seminar
From Jennifer Biddle April 16, 2021
11 plays
11
0 comments
0
You unliked the media.