A Conversation with David Smith
At CoreLogic, one of the areas we keep close tabs on is how weather and climate affect property as this could have some surprising downstream effects. The CoreLogic Hazard HQ team keeps a close eye on the state of weather covering catastrophes in the US and around the world.
Host Maiclaire Bolton Smith sits down with David Smith, Senior Leader of Science and Analytics to answer some questions about what is a loss estimate? What exactly are we estimating? Why are reconstruction costs different from actual lost?
MAICLAIRE BOLTON SMITH: Welcome back to Core Conversations, a CoreLogic podcast. I am your host, Maiclaire Bolton Smith, and I’m the senior leader of research and content strategy with CoreLogic. In this podcast, we’ll have conversations with industry experts about key topics, from housing affordability to the impacts of natural disasters on property. At CoreLogic, one of the areas we keep close tabs on is how weather and climate affect property as this could have some surprising downstream effects. From homes being destroyed to the impacts of construction material, demand surge to long-term ramifications on loans and home prices in the area. Natural hazards have a way of upending life as we know it. The CoreLogic Hazard HQ team keeps a close eye on the state of weather covering catastrophes in the US and even around the world. All of which, if you haven’t checked it out yet, you can find at hazardhq.com.
The Hazard HQ team filled with scientists. Many of whom have appearances on this podcast already release up-to-date analysis about ongoing hazard activity. And sometimes this can include one of the most discussed and most coveted piece of information, a financial loss estimate. For instance, in 2017, CoreLogic estimated that flood damage in Texas alone from Hurricane Harvey was estimated between 40 and $59 billion. Or in 2018 when the campfire broke a record for the most destructive wildfire in California history, CoreLogic estimated total losses between 11 and $13 billion. These numbers certainly are staggering. And because of it, they’ve been splashed across TV nightly news, but what do they really mean? And how do we come up with them? So today we’re talking with David Smith, senior leader of science and analytics to answer this complex question. David, welcome to Core Conversations.
DAVID SMITH: Thanks, Maiclaire. Great to be on the podcast. Thanks for having me on.
MBS: Awesome. Okay. So to get started today, why don’t you start by telling our listeners a little bit about yourself and your role here at CoreLogic.
DS: Sure, as you mentioned, I’m one of the senior leaders in the science and analytics group at CoreLogic. My role, in particular, is in terms of building catastrophe risk models. Catastrophe risk models are designed to quantify risk arising from natural hazard events, such as earthquakes, hurricanes, floods, wildfires, and so on. And we do this in the US and around the world. Building that brings together a remarkable variety of expertise. You need earthquake scientists, weather scientists, hydrologists, wildfire experts, structural engineers, to tell you how different types of buildings will perform, statisticians, actuaries, software experts, and so on. So I get to work with all these amazing people. And as a team, we build some really interesting and useful models, and I can even absorb a little of their knowledge along the way. So my academic background is in physical science, in physics and math, and in atmosphere geophysical sciences.
And this has provided a great foundation for my career. Formatively, some of my earliest interests growing up may have fed into this. I don’t want to oversell it necessarily, but I was, from early years, very fascinated by maps, and a big part of catastrophe risk is definitely the geography of risk. And I also had interests in high rise buildings and in the world around me. So I only made the connection to catastrophe risk as a career through a bit of luck in that fellow grad student was married to an investment banker who was on a flight and happened to be sitting next to one of the founders of EQE, which was at the time, one of the foremost companies looking at earthquake risk and they were looking to expand into other natural hazards. So that conversation, along with a brochure, too, was passed along to me. And that’s really how I got started in this very interesting field.
MBS: The rest is history. So well, I am really happy to have you on here, David. You and I have worked together for a number of years, so really excited to get your views on this important topic. And we’ve got some late-breaking information to you. That’s just unfolded in the last couple of weeks, too, that we’re going to get into. So before we get to that, we’ve talked about hurricanes, wildfires, tornadoes, earthquakes, extreme freeze, everything on this podcast so far, but we haven’t really talked much about event losses. So when a damaging event occurs, CoreLogic will release a loss estimate, which is a form of billions of dollars or millions or billions of dollars of this amount. But I want to unpack this a little bit. So we talk about loss estimate. We talk about X billion dollars happening in this event. What are we estimating?
DS: So when we, as CoreLogic estimate the losses from an event, what we’re estimating primarily are the costs to repair and rebuild afterwards. So from whatever damage that event has caused, our loss estimates are focused primarily on residential and commercial properties. And when we say commercial, we’re talking about things like industrial facilities, in addition to things like office buildings and shops and hotels and so on. And in this, we include all the costs, not only to repair, rebuild those properties, but also what it costs to replace or repair their contents, and also associated costs of their downtime. So things like business interruption for commercial properties, things like additional living expenses on the residential side. We also often break the loss estimates down, for example, by that residential versus commercial, we may look at geography by state, metro area, and so on.
And by specific peril. So hurricanes, for example, often come with varying impacts from wind and from storm surge and from rainfall-driven flood. And we might break those down, we generally do, by those different perils. We also often estimate the portion of losses that are insured, that is, will be covered by insurance companies versus the portion of losses that are uninsured will therefore usually be the responsibility of the property owner. So particularly for perils, like flood and earthquake, the uninsured portion of the loss can be very significant. For flood, it’s fairly common to see roughly 70% or so of the losses to be uninsured.
MBS: Yeah. I’m glad you just got into that because I think initially when people hear it’s X billion dollars, so let’s say an event is a $50 billion loss. People may just intuitively think it’s $50 billion to make the city new again, but that’s not at all really how it works. And I liked that you got into some of that. And one of the things that you just mentioned is that insured versus uninsured part. And I think, we’ve talked about that a little bit more on this podcast, in particular, when we’ve talked about flooding. And how flood losses, and we saw this dramatically in Hurricane Harvey, that there was so much of the flood loss was uninsured. I want to unpack that a little bit and talk about why things are uninsured. So I know on the flood side, we often see a really big gap between what the actual full uninsured losses versus the insured part of it. Whereas on wind, for a hurricane, for example, that’s a much smaller gap, and it’s because there’s a lot of insurance with wind and can you talk about that a little bit?
DS: Yeah, yeah, exactly. So this is a really important area and you’re already getting into the heart of it. I think, one answer is that a typical homeowner’s policy, at least as they’ve evolved over the last few decades, doesn’t include things like flood or earthquake. So that’s one of the biggest pieces. So such a policy does cover quite a bit from different kinds of natural perils, but often in general, it does not cover things like flood, earthquake, just the way the market has evolved over decades, last few decades. There is, with a flood for homeowners, if you have a mortgage, at least you’re required, if you’re within a special flood hazard area zone, to have that flood insurance. But if you’re not within, what’s basically effectively a 100 year flood zone, generally, there’s no requirement saying that you have to have flood cover. So it then goes back to the discretion of the homeowner and sort of the need for them to have an awareness of that risk and balancing that with the cost of the policy and all that.
MBS: Yeah. I think that is something we’ve talked about on this podcast a little bit as well, too, about how, if you’re mandated to have insurance, you get insurance, but if you aren’t mandated, you don’t necessarily think about it. So that’s where the awareness is such an important part. I think when we think of wind losses and wind being so heavily insured. There’s still a small part that is uninsured, is that just attributed to the deductible that people have to pay before their insurance kicks in, or what’s the attribution of that smaller little bit that’s maybe uninsured in the wind side?
DS: That’s right. So that’s the other part of it. So yeah, one part is, the properties that just aren’t covered for particular peril like floods, but that’s exactly right. For wind, what you can often have is yeah, very large deductibles or in some cases, limits may come into play, but where you have an event where you may have sort of widespread levels of, low levels of damage, you may have a big sort of gap between the insured and overall damage for that reason. Exactly along the lines, you’re getting in.
MBS: Okay. That’s helpful. Thank you, David. So, okay. The other comparison I want to get into is sometimes we also will talk about the reconstruction cost estimate for properties at risk. So, usually we look at this prior to landfall of a hurricane. So let’s say a hurricane is coming, we’ll say these are, there’s X number of properties at risk with a total reconstruction value of a trillion dollars. A lot of times it’s a really big number because there’s a lot of properties at risk because there’s a lot of uncertainty on where exactly the event’s going to hit. Can we talk about how that big number of reconstruction costs is different than the actual loss and why it’s so different?
DS: Right, yeah. So those are reconstruction cost values or RCVs. So individually they represent the cost to rebuild a building in the event of a total loss. So if it’s completely destroyed, what does it cost to rebuild it? And that’s right. We often do issue statements about sort of the total amount of RCV that’s potentially in harm’s way from an incoming event. And that’s really the key word, it’s potentially at risk. And so when, let’s say, a hurricane comes in, when it actually hits geographically, perhaps only a portion of that total area that we’ve looked at is really going to be significantly affected. And for sure, many of the properties aren’t going to be completely destroyed. So I think that’s really where we start to see that big difference between sort of the total reconstruction cost of what potentially is in harm’s way versus the actual damage and cost to repair it that occurs with the event.
MBS: Right, because we may not get a full loss of something. If there’s wind damage on something, it could be a partial loss versus a full loss in the RCV, or the reconstruction cost value is really looking at a full loss.
DS: Right. And the part, just to build on that a little bit, the part of the model that really gets into that is what we call the vulnerability model that really is having a lot of structural engineering behind it and research as well as data from prior events. And that really informs, for a given level of wind speed or flood depth and so on, how much of that RCV is it going to take, as a percentage of the RCV for a particular type of building for that level of hazard? What is it going to take to repair?
MBS: Yeah. And that triggers a thought to David that we probably, education for some of our listeners that may not be familiar with catastrophe modeling, that there’s several different components that make up a catastrophe model. You’ve mentioned the vulnerability model. We also have a hazard model. You want to talk just a little bit about the bits that come together on the modeling technology that we use behind getting these numbers that magically pop out?
DS: Yeah, sure, sure. That might be helpful. Yeah. So at the highest level, there are three main components. One is the hazard, the second being the vulnerability we just touched on, and the third is typically called the financial model. So the first hazard is really focused on the event itself. What kind of wind speeds does a hurricane bring with it? Very locally? It varies within the area affected. The vulnerability we’ve talked about, and ultimately that for a given type of building quantifies the financial cost to repair, replace, following on whatever damage occurred. And then the final piece, the financial model is bringing in things like the deductibles and limits we touched on a bit ago, sort of the insurance side of things and other aspects to sort of looking at uncertainty and really building out a full loss estimate, including quite a few different dimensions, but the highest level, those are really the three main components.
MBS: Okay. That’s really helpful context. I think David, as we get into some of this and I want to circle back to loss estimates again. Because we’ve got something really timely, that’s just happened in the last couple of weeks is, Hurricane Ida made landfall as a major hurricane, a strong category four hurricane on August 29th. Louisiana was pummeled again, poor Louisiana has been impacted so much by hurricanes in the last year or so. And we’ve talked about that as well on our podcast, too. Can we talk about the loss estimate for Hurricane Ida? And Ida was an interesting storm, and please get into this too, because the storm continued as it became, continued on as a tropical storm and it hit the Northeast and caused a lot of flooding. So I know we’ve had a couple of loss estimates for Ida. So can you just talk about what did we learn from Ida? What was interesting and unique about Ida as a storm and what do our loss estimates mean?
DS: Yeah, yeah. So, Ida came in Sunday morning, I guess, right around noon central time. I think it was, into Louisiana as a strong category four. So as you referenced there sort of different pieces to this, and really almost, not strictly true, but almost two distinct areas impacted by the event. So first we had really the three perils all at once in Louisiana and Mississippi primarily, and then some surrounding areas that were causing a lot of wind damage.
We also had some further substantial storm surge with it, as well as quite a bit of rainfall, particularly in the New Orleans area, and immediately west, fairly significant impact for sure, very major event from that point of view, a lot of consequence. And then a couple of days later, as Ida progressed north and east, up into Northeastern US, then we had an additional very major set of impacts from rainfall and then the flooding that ensued there. So really major up there as well. And so we’ve been scrambling pretty seriously since that Sunday, last week, really working on the loss estimates, trying to understand the event, trying to communicate our understanding of the event.
MBS: So if we look at those two kind of separate events, even though they’re both the same, Ida. They were very different events that happened from an insurance perspective, even. Can you talk through some of the losses that we had for Hurricane Ida in both parts of it?
DS: Sure. Yeah, because of the timing, we did issue our loss estimates in two phases. I think one, to sort of get the communication out there from the very serious impacts in the Gulf Coast area and then later, and quite recently we’ve issued our estimates for the Northeast. So in terms of overall property losses, and this includes, as we were talking about before, the insured part, as well as the uninsured in aggregate, about 27 to $40 billion of loss in the Gulf coastal area. And that’s from the combined effects of wind and storm surge and rainfall-driven flooding. So, very heavy 27 to 40 billion, a very major event, very catastrophic for many in the area. Of that, our estimate is for about 14 to 21 billion to be insured. So to be covered by insurance companies.
DS: We then went onto the Northeast, which is primarily a rainfall-driven flood portion of the event. And there, the numbers are very substantial as well. So things are still unfolding in a lot of ways, but there our estimates are for again, comprehensively, in terms of property damage, around 16 to $24 billion, of which about five to eight billion is insured. So very, very significant really in both areas.
MBS: Yeah, definitely very significant event. And I don’t want to overlook the human impact of this event as well. And it has had a substantial impact, with a number of lives lost as well, too. And from a scientific perspective, what we are trying to do is help understand the financial impact of an event like this. So that’s really the numbers that you were getting into. One thing David, that stood out in my mind as you were talking is, of the top of this episode, we talked a little bit about Hurricane Harvey and in previous podcasts, we’ve talked about Hurricane Harvey and how the real story there was, it being a major flood event that there was such a high proportion of the event was uninsured.
And some of the numbers that you just mentioned, it looks like, especially in the Louisiana Mississippi area, that the gap in insurance for flood was maybe a little bit less. And what’s the reason for that? When we looked at Texas, it was 70% uninsured. Have people learned since Hurricane Harvey that flood insurance is important, or what do we think is driving that being a smaller gap?
DS: That’s right. Yeah. So that definitely is one of the interesting pieces here. So instead of that 70% being insured, we’re seeing here something more like 50 to maybe 60% uninsured. So certainly substantially less on the flood side, primarily focused on the Louisiana area and different things. I mean, I think the most hopeful would be that yes, people are seeing the risk and acquiring coverage and that gap in coverage is getting smaller and I think that is part of it. Louisiana, has had some pretty significant events, including really major flooding five years ago, focus a little bit farther Westmore on the Baton Rouge area, but certainly would be close to mind. The other piece also is that a lot of this was in coastal areas, in areas that are really known for having a lot of flood risk and indeed having a lot of territory within those 100-year flood zones. So that’s sort of the other side of it that’s not necessarily as optimistic, but just reflecting.
MBS: Right. The mandate.
MBS: Yeah, no, that’s helpful. Thank you. I guess the other thing David, I think of is when we hear these numbers, a lot of times we’re hearing them on the news, but the media is not necessarily the number one priority when we’re creating these loss estimates or determining these loss estimates. Who uses these numbers and how do they use them?
DS: So, the audience for the loss numbers is pretty wide. I think there are a lot of different types of organizations and companies and so on that look to them. I mean, I think the closest to our world as CoreLogic is across three main types of company or entity. And that would be insurers. It would be mortgage lenders, and it would be the property owners themselves. So they all have an interest in as early as possible, getting a sense of what does this mean for them financially? So the insurers are keen to manage their risk and stay in business so they can continue to provide that coverage. And so the earlier they can get a handle on what this means. There are also aspects of what they need to do to react to the event in terms of the adjustment process and actually the mechanics of making the payments that provide that coverage. The mortgage lenders, similarly, want to manage their risk as well. They may be interested in knowing, for example, are there large numbers of homeowners that have been really severely impacted that may then have trouble paying their mortgage, for example.
MBS: Sure, yeah.
DS: So there’s a lot of reasons, and the property owners, of course, and this could be individual homeowners out to owners of portfolios of commercial property, they obviously have a very vested interest in this as well. And the general theme is everyone wants to really understand what does this look like as early as possible? And we’re doing everything we can to sort of help with that as the events unfold.
MBS: Yeah. No, thank you for that. And I guess, it leads me to think to, we talked a little bit about this when you talked about the components of what goes into building a catastrophe model, but how do we come up with these loss estimates? Do we just pop in the, this is the track of the hurricane. This is the properties at risk; it’s bang, this is the number. How do we come up with these loss estimates?
DS: Yeah, there’s actually a lot that goes into it. I mean, certainly having the model in place before the event is a key part of it. But of course, coming up with the specific representation of the event is a really big piece. And I think even beyond that, one of the key aspects of the data is what we call the exposure data itself. So a database essentially, of all of the properties throughout the region, be they residential or commercial, along with their characteristics, along with their location. And along with going back to that RCV, along with that reconstruction cost value. So there’s a lot of data that goes into it that basically needs to be run through the model once we’ve developed the representation of the event. And that is really ultimately what feeds into the loss estimate. In addition, there are things like the insurance conditions, [crosstalk 00:24:06] including, going back to what we were talking about earlier, the fact, whether a particular property is even covered at all for that peril, as well as if it is covered, what deductibles and limits it has.
And in terms of the representation of the event itself, this is really dependent on the type of event. I mean, I think, hurricane is sort of top of mind here. We’ve just had Hurricane Ida and so on-
DS: … but certainly, we do this for other types of events and it may play out differently if it’s an earthquake versus the hurricane wind side versus-
… flood. But generally it involves bringing in as much of the available data that we can, whether it’s overall information about the event, whether it’s in the case of hurricane wind, let’s say, specific wind speed observations throughout the footprint, so to speak, of the event.
To do our best, to really have that very sharp representation at a very high resolution of the event that we can then run against that database of properties. So the highest level, that’s sort of what we’re talking about-
… really, that property data, including very specific locations for it and the valuation, and then our representation of the event and its associated perils at a high resolution.
MBS: So it is a very complex task to really try and sort out. And I know a number of times immediately after the storm has hit, we start getting requests for what’s the loss estimate and they can’t be done instantly. It does take a lot of data crunching and running models and making sure and validating a lot of the storm footprint to make sure we can get that in. And I know with Ida in particular, we needed to wait for the rain to stop falling before we could start the loss estimate, because it may have looked like the event was over because the hurricane had passed, but there was still a lot of rain falling. So I think those are all important things that go into when you’re calculating this big number. There’s a lot of complexities and interdependencies that go into getting it. So…
DS: Yeah, that’s absolutely right.
MBS: Okay. So last question. I think just as we wrap up here. When people read about the events that have happened, they may come across different loss estimates from different companies. Why are they different? Why do different companies come up with different losses? And one company may say an event is five to $10 billion, and one may say seven to $15 billion. What do you think can be attributed to why different companies or different models come up with different estimates?
DS: That’s a great question. That’s probably on the minds of anyone that’s looked very much at loss estimates like this. I mean, I think a lot of times, first and foremost, you need to read carefully to make sure you’re seeing an apples-to-apples comparison. I think that’s probably the biggest thing. What exactly is being estimated? What’s being included and not included? We have this insured versus total loss dichotomy. We have which some perils are included? Are we including in them flood or aren’t we?
… and so on. So that’s, I think really the main thing to start out with, make sure that you’re comparing like for like. One other thing I’d say is that in the early days of an event, and I feel like there’s been more of this over the last few years then there was prior. There are sometimes some very high-level estimates put out by organizations that aren’t so much doing any modeling per se, but they’re really looking at top-level estimates, probably by referencing similar events that have occurred in the past and attempting to adjust for the differences in the event.
Again, the world is hungry for these estimates, right? Everyone wants to, as soon as they can, get a glimpse into what things are likely to look at for them financially. And so, that’s one way to meet the need. Those estimates that are really very high level, sometimes, they’re in the right ballpark, but they can often vary pretty wildly. I would just say that when we get back to the modeling we do. The better the data and the modeling that you have and that you bring together to produce these estimates, usually the closer you can be to the mark. And I would say sort of a caveat there is that there are events that sometimes occur that are just different enough from what’s happened before, that pretty much everybody trying to do these estimates may miss the mark, sometimes from the low side, and Hurricane Katrina in 2005 was one like that.
I think just the magnitude of what happened in New Orleans with Levees failure and just the enormous storm surge and everything was a challenge because it was just so different from what pretty much everyone had observed up to that point. The good news is that the data and the models are always getting better and we are learning from each event as we go on. So the likelihood of really big surprises, I do believe, is going down, not to say that, going to zero, but they are going down.
MBS: Yeah. And I think that’s probably one of my favorite things about what we do David, is that we are constantly learning, and every time another event happens, we learn something from it and it helps us refine what we do. So this has been so interesting, and timely, David, so thank you so much for coming on today to be a part of Core Conversations, a CoreLogic podcast.
DS: Thanks, Maiclaire. This has been great. I’m very glad to have been part of the podcast.
MBS: Awesome. Well, we’ll have to do this again, maybe when we have another damaging event, not that we hope for them, but it’s always good to add a little bit more insights into events after they happen because to help people understand. Because, as we say here at CoreLogic, to know your risk helps you to accelerate your recovery. So for more information on the property market and the housing economy, please visit us at corelogic.com/intelligence. And for information specifically about hazard events, please do visit us at hazardhq.com.
Thanks for listening. I hope you’ve enjoyed our latest episode. Please remember to leave us a review and let us know your thoughts and subscribe wherever you get your podcast to be notified when new episodes are released and thanks to the team for helping bring this podcast to life; producer Rhea Turakhia editor and sound engineer, Romie Aromin, and our social media team. Tune in next time for another Core Conversations.
©2020 CoreLogic, Inc. , All rights reserved.