Hi, welcome to today's webinar, Let the Cats Out of the Bag, a lesson on natural catastrophes and modeling, hosted by Inland Marine Underwriters Association. I'm Eileen Monreale, education training specialist at AIMU, and I want to ensure you that no animals were harmed in this production. Thank you for joining us today, and thank you to Lillian Colson and Eileen Monreale for their work to ensure the success of today's virtual event. On your control panel, you'll see the attendee control panel where you can see who else is in the meeting. We have a great turnout today, and I'm sure it's climbing as I'm speaking, because we're all last minute hopping into meetings. That is also where you will find the handouts section, and you can download the PVF of the presentation. And most importantly, it's where you can find the question box to share your questions with our presenter today. If that control panel minimizes by default, simply click the orange arrow to bring it back, and you can also go into the view menu of the control panel and uncheck the auto-hide option so it doesn't continue to do that. Our presenter is very happy to address your questions today. If they are relevant at the time that he's presenting, if they're shared with us, I will pose the questions to him right then and there. I'll take them throughout, and otherwise, I will wait until the end of the presentation. So feel free to enter your questions at any time. As always, to encourage your engagement and to show attendee appreciation, I anyway will award a $25 Amazon e-gift card to the most slash best engaged as a token of our appreciation. So without further delay, let me move my slide. There we go. I went too many. Without further delay, I'm honored to introduce Howard Kunst. Howard is a senior modeler and chief actuary at CoreLogic, and his credentials include certified catastrophe risk management professional, fellow of the Casualty Actuarial Society, and member of the American Academy of Actuaries. Howard is no stranger to the insurance industry and spent 28 years in our world, including as an actuary at Zurich Insurance and Union Standard, where he was VP and chief actuary. Howard began at CoreLogic in 2012, and when I asked him what he most enjoys about his work, he shared, I work with a lot of very smart people. We have a bunch of PhD scientists that I have learned a lot from. In his spare time, Howard enjoys golfing and exercising, and we're delighted to have him with us today. Please welcome Howard Kunst. The floor is all yours, Howard. Thank you, Eileen. Thank you for giving me this opportunity to chat with you all here today to talk a little about catastrophe modeling and what we're doing. As climate is becoming much more front and center, the modeling and even regulation, which I'm not going to really talk a whole lot about, but the regulators are very interested in this topic as well. We do a lot of work across the entire industry in helping educate people about catastrophe models and how they're used and the benefits that we get from catastrophe model. Again, I say this all the time, the biggest impact and why we use catastrophe models is that the historical record by itself just is not sufficient to provide enough information to use. Catastrophe models fills in the gaps of where history has lapsed. This is my agenda. I'm going to talk a little bit about some different types of catastrophe models and how they're used. Then I'm going to get into a couple of the actual natural catastrophe perils that probably impact in the marine underwriters more than some of the others and show you some statistics and just look at what's going on with those different perils, and both visually and just give you a bit of an outline of what's going on. I'm going to start with an overview of catastrophe models. I think in the market and many of these models are being used by probably some of you in the audience. We talk about the different types of models. We start with deterministic models and they are – let's get into that. This is like a score model. We have probabilistic models that provide AALs and PMLs and all different types of metrics around what can happen. The one that people are probably less familiar with are forensic models, and I'll talk about that here in a minute. Then we'll talk about how these models are used primarily relative to underwriting pricing risk management where there's overlap and how they all work together to provide an answer. Again, the types of models. On the left here we start with deterministic models. These can be models that could be what is the frequency of an event at a location, what is the relative loss cost at a location, but it's a score that represents some risk measurement and the higher the score would equate to the higher the risk. Again, based on what we know about the science, usually quite often these are like what's the probability of an event that could cause damage at a location and then we rate them across the entire – you could do it by a state but quite often we do it by the entire country is rated based on that same scale. Probabilistic model. These are the models that have been around for a while that are primarily used in risk management and reinsurance. Again, here's where especially the historical record isn't sufficient, so we will have historical events plus many, many simulated events that are used. Then there's a – I just simplify it. I'll say there's a Monte Carlo simulation where you do a number of simulations. All the modelers do different numbers between like 20,000 and 300,000 simulations of what could happen next year. The first thing you do is simulate the number of events in the year and then the second thing is given that number which are the events from the event sets that would be in that year and if you do it 300,000 or even 50,000 times you get a good representation of what can happen and especially as you're really trying to understand those tail events. The third one here that again is less known about are these forensic models. These are models that will tell you what just did happen. I'll show you some examples of that here in a bit but basically with the advanced technology radar based and ground based truth within a very short time the models can look at the radar and see what happened and predict both how big and where hail fell, what were the wind speeds at various locations. Again this information can be used in both underwriting and claims to help people understand some of the risk here. These are the main models here. Whoops, I don't know what happened there. I don't know how I got out of whack here. It's not going. Sorry, it went too far. Back a slide? Is this where you need to be? Please. So you want to be on slide five or six? I want to be on slide five. Okay, I think you are now. I don't see. There you are. Okay, yeah, so you're on slide five. There we go. So again one of the things I was going to add here is that with these probabilistic or with the forensic models because we can see the various events as they occur and see the footprints of them in addition to providing information on where these events can happen we do get a good variety and a good amount of information on what these events look like, what their actual footprints are and as such those footprints are then used to help inform our probabilistic model as we create events that go into the probabilistic models when we create the event sets. Again these event sets are built on the knowledge of the signs of a particular peril. For hurricanes it's based on how we know hurricanes move and how they grow, how they impact with the sea surfaces and the sea temperatures and a number of other things. So when we actually have then the real events to go with that we can build some pretty robust event sets that can be used for these probabilistic models. Now there you go. So now we have a poll and I don't know if how this is going to work but here we go. There's a poll again which one of these bottles is used to help inform one of the others and please please vote. So, 40 percent of you got it right, it is forensic models that are used to help inform the probabilistic models. They also do help as we're developing the, you know, the, what our expected frequencies are at various places. They can be used in the, how we determine some of the deterministic models as well, but they're, the, but forensic models are the ones that are used to help inform the other models. So, so thank you there, move on then. So let's just talk about some of these models again, a little more depth. Again, the first one I discussed were deterministic models. And again, these are models that represent a relative risk for a specific peril. For example, this map here on the right is a map from Houston. It's showing the flood score. As you would imagine, you can see in the red there, there's probably a river that's going through there. And part of this is, is, you know, you can visualize where the, where the risk is. As well as we can show the, there's more than just the, you know, the flood zones that are in, that take into consideration, it's really based on elevations. And you can see in the colors there, you can see where the yellow go extend, you know, beyond just nearby the red in some areas where the elevations stay low after a while. So there really ends up being, you know, very, very helpful in understanding where the risk is. And the visualization of these models, you know, a score could be visualized very easy on a map. And it really helps pinpoint if you have one or even a portfolio of locations that you're looking at that are in, you know, nearby each other, you can really look at the risk, relative risks for the various locations. And probabilistic models, I mentioned these briefly. But again, it starts with understanding the hazard and the science. And this is where the scientists come into to play and understand what are the types of things that can happen at, from these types of events. This is a kind of a flow chart of, for an earthquake model, where you start out with the event catalog and based on what we know about from the USGS systems, we know where the faults are, what are the types of events that the geological survey thinks can happen. And so we build events based on the, you know, various types of events that could occur from those various fault lines. Same thing I mentioned earlier with hurricanes, you can build various events based on how hurricanes move across the Atlantic and then once they go inland, what the frequency of them are, you know, where they're going to hit and all these different things. And part of the, and how these models are used then is everywhere along the path of that event, we have metrics on how big the wind field is, how big the damage field is, and understanding what the event characteristics are at any location. So that knowing those event characteristics, you could then take that and turn it into the hazard at a location. So if event X, you know, impacted a location, what was actually the metrics of that event at a location? And then we can determine what the loss of that structure would be based on, you know, the structure and all the secondary building characteristics and information like that that would determine how that structure would perform under those situations of the event that is being modeled. Again, there's many things that go into that and, you know, we'll talk about a little bit later, but all the, this is where we can really measure the impacts of mitigations and the things that you can do to reduce your loss, which in these times is becoming much more and more important and the resiliency of structures. So this is how we can measure that and we're working with a number of clients doing that right now. And then once you get the, all those ground up losses that can happen, then you apply financial terms, you know, limits, deductibles, you know, a lot of other factors to understand what the actual insured loss might be. With this, you can do a number of outputs that include the, you know, the, for any structure or for a portfolio, you can look at what is the loss from any event? What is the loss in any year? There's a lot of different tables that you can use and that will help you understand then what those average annual losses are, which is the AL and then the PML, which is the probable maximum loss. What are the types of things that will happen once every 100 years, once every 250 years, and also then kind of the variation around the mean of what can happen and a lot of that is used in both pricing and in determining capital management and reinsurance. Again, forensic models, here's an example of something that happened a few years back in Minneapolis, a hail event. These models are very granular, they're down to, you know, hundreds of yards, sometimes even closer than that for understanding where the event occurred. These are particularly helpful when you see a map like this and you can plot a map like this overlay over the top of a number of assets that a company may have, whether it be an insured portfolio or even, we've done this for looking at the impacts to big restaurant chains that are in metros that have been hit by these events. We can tell them which of their assets were impacted and depending on what we know about those structures, we can also give a preliminary impact of what the loss was. I wish I had this type of information handy when I was in the business because I was, as Ellie mentioned, I was chief actuary at Union Standard, which is a W.R. Berkeley company here in, headquartered here in the Texas area, and any time we would have a wind loss or hail loss in the states that we did business in, the next morning the president would be at my office knocking on my door trying to get a feel for what we thought the losses would have been and this type of information is really critical in helping with at least a first glance of what the loss could be based on knowing how big the hail was at any location. And you can then also triage where you send your claims adjusters. So again, these models are all used together across a suite of uses. When you talk about the insurance activities that you guys are doing, it starts with screening and underwriting if you want to take the risk or not, what are your risk appetites for these various hazards. Right now we're seeing a lot of that in, that's an issue with California wildfire and I'll talk a little bit about that later, just understanding the affordability versus availability issues that come from that, that the homeowners in particular market is having a big issue with right now. And as you take those models, if you trust them in the use of screening, the next step is to use those models then again for pricing and again you can use, if the scores are telling you where the risk differentials are, you can use those risk differentials from the models to understand how much different the charge rates, the rate charges, the premiums for the various risks and such we can build those into. And they're really easy to build into a rating algorithm, putting in a factor that is for individual perils, especially if you're doing biperil rating, you will put it for that particular peril, you'll base rate times a relativity factor based on the score and also based on the building, the construction type of a structure. When you get into more complex the structures, industrial structures, agricultural, larger structures, you also use the probabilistic model to help with that because they will take into consideration some of the building characteristics, some of the mitigation, just what is the structure more likely to, based on what's actually there, give you a better feel for the losses that can occur as opposed to just the general risk because again, as you get into these larger structure, they get to be much more complex. And of course, when you talk about portfolio risk and underwriting, understanding your concentration of risk, you can use the score models as well in understanding that if you run your entire portfolio and assign the risk scores to them, you can map it and see if you have concentrations of risks, of higher risk in certain areas. And again, and then validate that with the probabilistic model and look at the events that have the largest dollars of loss and where those are to help you understand if there needs to be some additional underwriting of the book to minimize potential losses. And again, we're seeing now in many, in a number of pearls where companies are using these models to help with exactly that is to manage their aggregates because it's really, as climate risk is growing, they're really trying to minimize the impact of any single loss event into the book. So using the models for portfolio risk and understanding concentrations and aggregation of risks is something that's been very front and center over the last few years where these models are being used more and more. So let's talk about flood risk. So flood, there are a couple of different versions of flood. Along the coast, you have the coastal storm surge. You have just the general everyday rise and fall of tides. But as hurricanes roll in, they push water in front of them and bring extra water inland. We'll talk about that. And of course, there's inland flooding, which is due to rivers and lakes and streams rising from events. And also flash flood, even though there's no rivers or lakes around. Flash floods when there's been an extreme rainfall event, where is it that water can gather that will create floods in a very short period of time? So storm surge, again, I mentioned briefly. Now we didn't go too far. Let me go back. There we go. So the coastline is very, very different based on the bathymetry of the ocean bed will impact how much water can flow through, the size of the storm, and the intensity of the winds is also part of that. And so depending on what the winds are and the size of the storm, we will know what can be pushed. And then based on the bathymetry, we can tell at any location how large a wave event can you expect from a Category 1 hurricane up through a Category 5. One of the interesting things here is where you have very shallow coastlines, there's no place for the water to move once it starts moving through. And so it all just keeps getting pushing and gathering. If you have deeper water, the water can push, but there's still then the water can escape back underneath and back out into the behind it. But generally speaking, shallower coastlines produce these larger waves of events. And we see that every time when a hurricane event hits the Tampa Bay area, because Tampa Bay is very shallow, it ends up pushing all that water into the inland. And there's been times where the Bay has been almost dry because of it. And so that's kind of a validation that that's the impacts from a storm surge. But we see a lot of this. And again, it's really based on the water in from these events. And anything along the coastline that's at lower elevations is at risk. And I'll show you here in this next picture, here's a map of storm surge. And you can see that the areas of red are the higher risks, and it does go quite a bit inland. Again, one of the things that I was talking with Eileen as I was prepping for this is one of the important thing here is understanding what the elevations are. And you can see from this picture that there are areas that are right along the coast that have low risk. And that's because they're at higher elevations. So when it comes to coastal flooding, distance to coast is not the best solution. Understanding the elevations and having a coastal storm surge score and mapping it, like I mentioned earlier with some of these scores, really helps you understand where the risk is and where you may want to be more aggressive riding business. So here, if you're worried about storm surge, there's a pocket there in the actually a couple of pockets there near the coast that you can ride safely relative to storm surge. Now, you still have the impact of hurricane winds, but you would be less of an impact relative to the flood risk at this location. And which is important given a lot of in the marine warehouses as the shipping comes in from the ocean goes into warehouses, a lot of those warehouses are near water. So understanding where the risk is really important and which warehouses are elevated, which ones are not. So we have another poll. So which data value is more important when determining risk for coastal flooding, distance to coast or elevation? Most of you are paying attention. Elevation is the most important thing. Higher elevations are less likely to be flooded, everything else being equal. So elevation is really the most important value there. Howard, we do have a question from Caroline Gannon. The question is, would mudslides fall into this category and further is flooding as well? So mudslide is actually considered earth movement, which would be earthquake coverage. So mudslides would be more earthquake coverage. However, there have been situations where it is earth movement, but when you have a wildfire that has removed all the vegetation along a slope, we have seen mudslide events in California, for example, that were ultimately allocated to being a covered loss because the praximate cause was the wildfire and it really wasn't earth movement or just a mudslide by itself. So mudslides are one of those unique things, but it's usually earth movement unless it's the praximate cause of the wildfire. Good question. And actually another one just popped in from Ben Wyatt. Do we have any data regarding the impact of tides conditions impacting? I don't know if there's really anything that we would have on that. I know we're seeing more and more as we see water surface elevations rising, global sea rise. I don't have any of that information. I think when we build our models, one of the things we do is we look at where the sea levels are right now. And so we're now looking at where it was 10 years ago or 20 years ago. We're building in the current values of all those elevations. And so it's being accounted for in the models kind of indirectly. But we do in the development of the surge models, for example, we do look at we have events that are low tide and events that are high tide. And we do know that it does have an impact in how much water can be pushed inland, depending on where the tide is. Any other ones? Actually yes, but would you rather me hold them or just those two more? Let's hold them. We'll get to them at the end of here. Perfect. Perfect. I agree. If you go to the next slide here, being a little slow here, there we go. So riverine flooding, again, our model will take into consideration where the FEMA flood zones are, but again, when you talk about FEMA and the flood zones, they're very binary. This is a line, you're in or out, and it's not as, doesn't provide as much information as a much more granular model. Flood risk score can go from like 1 to 100. Here you can see a map of Cedar Rapids on the flood event that happened about 10 years ago, maybe a little over 10 years ago. The blue lines in here are the FEMA flood zones you can see on, especially on the, around the water there, the AE zones, which are the 100-year zones. And then the X500 zones in there are outlined in the blue, and the red around the outside is the actually inundation area. And you can see in areas on the south and west sides of this event primarily, which should be in the bottom and west side of the maps, but there are significant areas where this particular event flooded considerably beyond the 500-year flood zone. Now that's the 500-year flood zone as FEMA defined it. But when you're looking at the actual modeling, the risk and the map, you can see that the CoreLogic model actually identified those areas as being high risk. You can see all the orange that is in those areas that I mentioned that were outside the FEMA, even 500-year zone, where the model had identified was low elevation and subject to significant risk. And so those areas were labeled as high risk. And so this shows how models can be used beyond what the, just the FEMA flood zones are. And again, the scary thing here, again, is that people are required only to have flood insurance if they're within the 100-year flood zone and have a federally-backed mortgage. But you can see that the risk really does extend much farther beyond what the FEMA flood zones are saying. Flash flood, this is really interesting. When you look at these models, I mean, again, elevation, as you would expect, is very important. When you have these events, water flows down to the lowest point. And you can see in this map, again, how the various watersheds all work together and where the water flows to the lowest spots in the areas. And you can see overlay on this map, I think, where the floods were based on and then where these low areas are. And it matched up pretty well. Again, the flash flood risk is a combination of the elevations and then, again, which areas have a higher likelihood of having excess rain, which is more than two inches of rain across a five-hour period, I believe is what the general assessment is for excess rain. And so, areas that can have excess rain combined with low elevations are going to be considered very high risk relative to flash flooding. Again, there's no lakes or streams or rivers associated with this. These are just pure rainwater flowing down to the lowest point. And these models, again, are very granular. The elevation models that are being used are at 10 meter by 10 meters, somewhere along, sometimes along the coast for storm surge, we're down to one meter. Because of how flood models work, they're very computationally intensive. When I started CoreLogic, the flood grids were 30 by 30 meters, then they went to 10 meter by 10 meters, which just increased the number of grids by a factor of nine. And so, one of the issues of why it took flood models a bit of time to be developed was that we just didn't have the computing power until 15, 20 years ago to really start looking at the models that we can build to understand where water is going to flow. So, convective storm, again, severe convective storm, these are the events that include hail and wind. And this is both tornado winds and the straight line winds. You hear them called derricos and other names of them, but those are the extreme winds. But these all come from the same types of events. Interestingly enough, the distribution and the risk from hail has a different view, if you map it, looks different than wind and tornado, as they're just a little bit different. All those events occur from these convective storm, thunderstorm events. And so, it's really interesting to show where that risk is and understand the change in risk. And so, this is really interesting. We've got these maps and you can see the different risk levels from the blue, which really is like 30% of the US, down to the red, which is like really 10% of the United States. And then overlaying where these events occur and we get questions all the time when you have a random event, like looking there in 2021, you saw a couple of events in Michigan, for example, and people would say, well, you said that was low risk, why is there a loss there? And again, the frequency of loss is much lower in the lower risk areas. You will occasionally have some losses in those areas, but if you look at the total number of losses compared to the landmass area of that risk description category, you will see that, and these are pretty consistent, that the frequencies per landmass area go up significantly as the risk increases. And so, again, you can see the majority of the events are in the orange and brown and red there, which are the high, very high extreme areas, which ends up being about 25 or 30% of the actual locations across the United States. Again, this is 2021, you can see the impact from hail. The next one I have is 2022, and you can see there was a bit more events, there's a little more dark black on here, these events, and again, the majority of them are in the high risk areas, but you have a couple of stragglers out there in the West, in Oregon, Idaho, Washington, a little bit there in Montana, in the low, but very low, there are still some events. But again, there are few events, and relative to the entire US landmass of that description, these are relatively small. Now we know that last year, there was a, you know, if we talk about the FEMA billion-dollar events or NOAA billion-dollar events, that last year, 2023, was really bad risk relative to the hail, and this map really does show it compared to those prior to. All the events that occurred last year for convective storm, all the hail losses that we saw, I mean, last year, the convective storm losses were driven by hail. Interestingly enough, this year, we're seeing more tornadic events, and so the losses from convective storms are leaning more towards wind and tornadoes this year compared to 2023, when we had all these hail events, as you can see on this map. So this really does show, you know, again, we all know that there were more, I mean, I felt in all the losses that we had last year from hail events, and this map really does show a really good job of that, you know, that there were a lot more of these events. And they're pretty much all in the really high-risk categories, just a few in the green, which is the low, and I don't know if there was even any in the blue, very low category. When we look over a number of years, again, with our – and again, those hail paths come from our weather forensic information, where we monitor where events – damaging wind events – the hail events occurred, and again, those were two-inch hail events or greater. One of the things we show here is, over the last 20 years, what has happened with extreme hail and wind? And if you look at the blue scatter line on top, that is the number of days in the year that had 90-mile-an-hour or greater winds. And if you look at the dotted blue line with it, you can see that actually the number of extreme wind days has slightly gone down, interestingly enough. However, when you look at the orange dotted line, the scattered line, which is the number of two-inch hail days, and then the linear plot line going through it, you can see that over the 20-year period or so, we went from about 110 days in the year of two-inch or greater hail to nearly 130. So there's definitely been, over the last 20 years or so, an increase in the number of days of the year where we've seen two-inch or greater hail. And I remember last year, we get these daily reports of all the locations that had hail, and there was one point in the spring last year where I remember three weeks – at least three weeks of days in a row – where some place in the U.S. had three-inch or greater hail. And you think about three-inch or greater hail as being pretty extreme. We had 21 or more days in a row where some place had three-inch or greater hail. That's just remarkable, and all the damage that was done. And I know – and in some of the other studies I've done and presentation I've done, that has impacted insurance companies' financials. A lot of insurance companies in the Midwest have reinsurance for catastrophes, but there were so many of these events spread across the entire year that, unfortunately, the reinsurance recovery was very small. So they spent a lot of money in reinsurance, but then never recovered as much as they put out in their hail losses. And then this year, as the reinsurers started trusting miles and did a bit more work to it, the reinsurers ended up cutting capacity. And so a lot of actuaries from insurance companies I've talked to have told me about the struggles they're having acquiring early-in-the-year cover for hail and convective storm reinsurance. And I'm sure some of you have experienced the same thing. So that's an interesting – just kind of looking at convective storms here and what's been happening. The last one I'm going to talk about is wildfire. Again, there's just so much going on with wildfire, and I know my colleague Dr. Tom Jeffrey has presented to this group on a number of occasions talking about wildfire. But again, this is such a growing topic in the West and even other areas across the United States where more and more – we're seeing more and more wildfires occurring. The impact of climate change is having some impact here as well. So it's really a risk that even in the marine underwriters will have some exposure to, especially when we're talking about builder's risk in the building, and even agriculture. We've seen a lot of buildings, especially in California and some of those areas that are having risk to wildfires. And I just had somebody in our office a couple weeks ago that I was talking about. They help counties build plans around disaster recovery, and they're working in Napa Valley with a couple of the municipalities there to help them build a plan to work with their wineries and the grape growers to understand how to protect both the grapes growing and then their structures at the wineries to keep them from burning, because they did have such a big impact just a few years ago in those Napa fires. Again, this is really a – you've probably seen this before. When we talk about wildfires, just to show you a little bit about the complexity of what some of these models are, I'm going to use a wildfire here to really dig in. You've got – obviously, the vegetation is really important, how dense it is and how combustible it is. The vegetation class tells us a little bit about whether there's been burns there in the past. There are certain types of vegetation that will only exist if there have been intense fires or intense heat in the past. Again, slope and aspect, which direction is the hill facing, how steep it is. Steeper slopes will burn faster. And those four things go into the assessment of what can burn. When you talk about the actual wildfire risk score, then you talk about what's near the property. You don't have to be right in a high-risk property in a high-risk location to still have high exposure to a wildfire, because you do have wind. Again, one of the things that we've seen over the last few years relative to wildfires is the influence of wind has become much greater. A lot of these large events have been wind-driven. We talk about the big Santa Rosa – the Tubbs Fire there in Northern California. It was a wind event. The wind blew embers over a six-lane highway and started a couple of developments on fire. These structures end up becoming fuel for their neighboring homes. Then we had these things called urban conflagration, where they just spread from home to home to home. I was actually speaking to a group of California underwriters just six months after that Tubbs event. One of the underwriters was telling me that his in-laws were in the Tubbs, Santa Rosa area, and their neighbors got them out of their house at 2 a.m. in the morning and said, well, the fire wasn't to their house yet. They could see it moving. It was moving that fast from the winds. There were structures in that event that were 2 1⁄2 or 3 miles away from any combustible vegetation, and yet still got destroyed in that event. Wind and embers flowing is really important. Just because you don't have vegetation on your property, you still may have some risk. I'm doing a presentation in October at the American Academy of Actuaries fall meeting about the impacts of mitigation. I was talking in prep with Dr. Ian Giamontico from IVHS, and we were talking about the level of mitigation that you could do if you're right in the middle of vegetation. They actually had some structures in one of the recent fires in California, and the name of it has lost me right now. They're actually showing that there's only so much you can do if you're right in the middle of that high risk. You can't do defensible space. You can maybe minimize it based on your building composition, but there's only so much you can do. But again, distance to the risk is really important here based on wind and embers and some of the other things that are occurring here. Again we look at the statistics here for wildfire. Interestingly enough, if you look at the chart in the top right, if you look at the total number of US fires over the past 20 years, 20 plus years, you can see there's actually a bit of a decrease. Part of that is that we've become much better at suppressing small fires, and so the number of fires are going down just because we're suppressing more fires early on. But when you look at the total number of acres burned, you don't see a decrease. It's kind of up and down. I believe there probably is a slight trend towards more acres being burned on that even though the fires are less. But when you go to the left-hand side, the actual structure is destroyed. Again, you can see the impacts of those 2017 and 2018 fires in the West, primarily California. 2019 was just a really odd year. There was hardly anything that happened. And then 2020, again, we've seen a bit more, saw another high last year. And then the last three years were again back kind of higher than we had seen in the years prior to 2017 and 18. One of the things we hear in some of the consumer groups, and it was brought up a number of times in California in some of the discussions around the use of modeling and the regulations that the California Department have been putting in, that insurance companies have made money over the last few years writing property in the state of California. And so why should they be charging higher rates? Well, I mean, the losses from 2017 and 2018 alone wiped out the last 20 years of profits of any insurance company in the state of California. So you really need to look at the long-term and insurance companies, I mean, homeowners are seeing an extreme amount of, as I mentioned earlier, relative to the affordability of insurance and the availability of insurance because of the wildfire risk. And until insurance companies can get rates to be adequate, we're still going to see some struggles there with the insurance market in California. Now, I mentioned, I was talking earlier about wildfires and mitigation, again, CDI put in a number of, they put in 12 mitigation items that insurance companies had to provide a rate credit for if they had them in place. It was a good first step, but there's more, there's many more things that can be done. And I think I left this at the end here to kind of talk about this in that as we see risk increasing in other areas of the country due to convective storms or hurricanes and flooding, we're going to have to be very proactive in the insurance world, working with the mortgage world as well that like CoreLogic works a lot with, and working together to do mitigation and to help doing whatever can be done to minimize the risk. Because if we don't, the expected losses are going to go out of control and the market will be unsustainable. So it's really going to be everyone working together from the consumers to insurance companies, to the mortgage companies, to the federal government providing the funds, which they do, for areas to do mitigation on various, for various different perils. A lot of it is around the flood peril, but there are things that can be done. And again, understanding that, what are the things that do have the biggest bang for the buck in reducing losses, it's really, really important. I work with a lot of, a couple of think tanks that are talking about how we can deploy some of these, this federal money and where we want to do this. And a lot of times they talk about, there was some study that done that for every dollar of mitigation that is spent, you can expect to reduce a loss by $6. Well, that's really important for the insurance company, but for you and me as a homeowner or as a buyer of insurance, if I'm going to spend $5,000 to do some mitigation on my structure, I want to know how much that's going to reduce my insurance premiums. And if it's only a couple of dollars, then I may not worry about it. But if it's a couple hundred dollars, then I may be more proactive in doing something to reduce my premium. So I think we really need to focus on the big bang items, the high ticket items that are going to have the most reduction in expected losses. I think, again, it's not only the insurance companies, but everybody as a whole working together to do this. And I think it's really important to understand this. And then again, with wildfires, we have all these things that are there. And again, there's not a lot of data on understanding exactly what all these are going to do. IDHS does a lot of work helping with understanding some of the research on this. But until we get more data and more information from exact losses, we'll be able to do a bit more on estimating the impacts of mitigation and resilience. So again, to kind of summarize my last slide, I think I talked about this, but granularity is important in getting information down to the structure level. And again, granularity in location is key as well. In some perils, 100 feet may not make a difference in the risk. But if you're talking about wildfire and flood, for example, 100 feet can make a huge difference in the risk for that location being impacted by an event. So granularity of models, granularity of the GIS and the geocoding that's going into the structure, structure level geocodes is really important. So knowing the risk onto that granular level is really important. And understanding what the building characteristics are as well to make sure you understand exactly what the risk is and what can be done to further reduce that risk. As I mentioned just now, resilience, climate change appears to be increasing losses. And we're seeing lots of demand across the country for mitigation to be implemented. And again, I think we need to figure out how to do that in a reasonable way to help minimize the insured impact and so that insurance stays sustainable and doesn't go out of control where only the rich are going to be able to afford insurance. So we need to really consider that, all the things that can be done there. And again, I mentioned this early on, historical record for catastrophe risk is pretty small. Some are just the last 50 years like wildfire. Hurricane, we have some information back 100, 200 years. But it's still not given the frequency of those events. That's really a drop in the bucket. So we need models to really help us understand what really can happen beyond just what was, what has happened. And again, when we talk about when catastrophe models really started becoming more popular was after Hurricane Andrew and the Northridge earthquake. Those events, both were events of a magnitude that were way beyond what anybody ever thought could happen. And so if we had catastrophe models that could have modeled the types of events that could happen, those events would have been in there and they would have understood that there was the possibility those types of events exist. And again, now knowing that information allows insurance companies to make better decisions around underwriting, concentration of risks, aggregations, pricing, and then what they can do about risk management and reinsurance. So on that note, that is the end of my presentation. And we have about 15 minutes for additional questions. That's fantastic. Thank you so much, Howard. As a reminder, just type your questions in and Howard will be happy to respond. So we have a couple of questions here. The first one is from Judy, how often is CoreLogic data updated for flooding? Good question. So CoreLogic actually works very closely with FEMA and we actually digitize all the flood maps for them as they come in, as all the changes come in. So in our flood risk score, some of that is calculated on the fly. It's like accurate up to like the minute. Some of the probabilistic models and some of the other models, we update that every year or two based on where and what information we get. And the other models, like say wildfire, because of changing vegetation, that model is updated quarterly. So it really does vary based on the model, but it's really key to the point of the question, keeping the models up to date with the current climate and the current ground truth is really important so that you're estimating, doing a better job estimating what can happen now and not relying on where things were five or 10 years ago. And we have some competitors that are not as frequent with their updates in the inputs. And I believe as an actuary, it's really important to stay up to date with the inputs. Thank you. So this is from Karen, and I'm not sure if your answer was buried somewhere in what you were just responding. But her question is, do the models consider climate change and where flooding is likely to occur in the near future, say the next five or 10 years? So generally speaking, the models are trying to predict what's going to happen next year. And again, it's staying up to date with the current climate conditions so that we're looking at where we, based on what we know is there today, where the losses agreed to be. We do have some work we are working on for climate risk analytics that is taking some of the IPCC projections and turning them into, you know, then downscaling it into localities and understanding if different scenarios occur, what does that mean for five years from now, 10 years from now, 20 years from now? So we do some of that with our climate risk analytics. That's separate from the modeling itself in that in the insurance world, you know, it's starting to change. But previously, insurance world was more worried about what's going to happen next year and not worried about five or 10 years from now. Where we've seen the biggest push for this type of information is actually from our mortgage customers that are interested in the potential for them getting stuck holding the keys to a property five years from now, 10 years from now because of climate change. We have a question from John from Enga. How do models account for builder's risk given that actual exposure is stored at zero and increased over time and the policy data is on completed value? Good question. I mean, we would, I mean, the insurance company would need to understand what the risk to that structure is based on the various peril. Let's say it's hail. They would need to understand what the probability from our model, what the probability of hail would be. And then as the value is changing across the period for that builder's risk policy, you know, the exposure would be changing and they would have to calculate that based on the, you know, across the period of the exposure and what's actually at risk at any point in time. So, we would provide them information on what the risk is and then the insurance company would have to then turn that into how they're going to rate it. Excellent. And Judy asks, when writing wildfire for a builder's risk, besides clearing brush, what other factors should we be concerned with during the underwriting process? Also, besides a water source, what else would or could we ask for from the insurer? Well, again, making sure that the defensible space, whatever the parcel that the home is on is cleared is really important, especially like that first five feet and then the next, you know, 10 to 20 feet, making sure everything is clear around there and free of combustibles. The other thing is just, you know, how close are you to actual vegetation that's outside your parcel? If you're, if it's a new construction and the parcel is completely, you know, bulldozed over and there's no risk, but, you know, do the neighbors have risks? Is there trees from the neighbors that could impact it as well? So, the community level mitigation is really important. It's actually probably more important than the individual structure mitigation. But yeah, just make sure, especially during the construction period, that if you're going to put in any vegetation for the landscaping to make sure it's done, you know, at the end and not provide any potential for fires. Again, the other thing is too, not only like vegetation, but even landscaping, but like fencing. The Marshall Fire in Colorado a couple of years ago that happened around Christmas time. Even though it was grass, which normally is not hot enough to burn homes, what happened is the grass burned up into some fences and then the fences actually that it took over and these fences were all connected to homes. And so, those fences burned right up to the homes and then burned the homes. And a lot of it was the concentration of homes and these fences. And on a side note, we have an office that's not very far from where that fire occurred. And we did some fieldwork and I actually got to see some of the burn and some of the destruction from that fire. And there were people that they didn't lose their homes, but they lost their fences. And sure enough, they put wood fences right back up where they burned, you know, just a month or two before. And I just couldn't understand why you would put a wood fence back up after you almost lost your house. Sure. But it happens over and over. They keep rebuilding here on Staten Island, despite the flooding, so it doesn't make sense. So here's a question from Mark Kane, referring to the four presented elements of the wildfire analysis, not including relative moisture levels of the flora and their surrounding soils. Is there any discussion of porologic using dynamic real-time perimeters that would reflect the conditions at the moment utilizing satellite information? So we've looked at that in our modeling and it's really hard to do daily risks. What we do look at, though, is like long-term drought and long-term wind changes. And that's a factor we didn't show in there, is the latest version of the model looks at a relative – adds to the score if the location has been having some significant drought over the last five years. And the same thing with relative to like an average amount of high winds – wind days. Is this a location that's in a high wind area or not? And so the risk is influenced – the score is influenced by those two as well. It's really hard – I mean, and again, if an insurance company is evaluating a model – use of model, if that model keeps changing every day, it makes it very hard. And it's really hard for our teams to do that dynamic modeling. So the wildfire – again, the wildfire things are updated quarterly in some areas. Everything is updated annually. So it's not done every day, but it's done every year to make sure that we're calling for the changes that are occurring on the ground relative to certain things. And again, there's more than just like the moisture of the vegetation, but there's also things like we know bark beetles have gotten into vegetation and caused those vegetation to die. And why that's important is that when that vegetation dies and it falls, it's on the ground and wildfires travel along the ground. It's not the canopies and the trees that are burning. It's the stuff along the ground. And so when you have dead forests, those are just ripe to have wildfires because it's additional ground. Thank you. Excellent. We have a question from Ben Wyatt again. Do the corologic models factor in inventory, contractors, equipment versus property? And how is damageability compared? So when we use the probabilistic model, yeah, I mean, inventory and contents is definitely included in it. And we, the model assumes, make some assumptions on the types of equipment and the type of contents and structures based on the occupancy. So there are some things that are done that try to understand, and that also works along with the reconstruction cost models that try to estimate the value of what might be in those buildings. So you have the reconstruction cost of the structure, and then you have an estimate of what the contents are. Now if that changes, I think that's at the insurance company level to take the information and understand that the loss cost for contents is a different number than the loss cost for the building itself. And then if there is more or less contents in there, and or contents that are even more combustible than average, they would need to modify that to get a better assessment of that risk and the long-term premiums. So we have another question from Karen. How are AI machine learning incorporated into models? Good question. I was wondering how long it's going to take. So AI and ML have been used for a long time in these models, and we just never really talk about it. For example, again, we get a lot of wildfire questions. We use satellite imagery to tell us what the vegetation is on the ground. And how is that done is that we take the spectral imagery from the satellite data and then compare it with what's actually known on the ground, and then we teach the model to look at what that imagery is, that the spectral imagery is, and teach it over a period of time to understand what the vegetation is on the ground. So there's a lot of different things like that. Again, understanding what if we can determine, again, for wildfire, for modeling, if there's five feet of defensible space around the house, what types of... Does it have enclosed eaves? Does it have a combustible roof? Things like that that can be learned from models that take the imagery and turn it into a... Have been taught as to what it means for the structure. I think one of the interesting things that can be used in the future, and I mentioned this briefly with taking the forensic information on what some of these footprints of actual events are, I think with the capability of AI and actually looking at past events, we can probably vary them and come up with a model, an AI model that will provide even better, more realistic footprints for events in the future that will be much more aligned with what could happen and where it can happen. So I think AI can be used more, but it's already been used in a lot of the understanding of what the variables are and understanding the risk already. So they've been used for a long time. Excellent. Thank you. And here's our final question as we're coming up on time. While fire losses have been increasing over the last decade, is that mainly climate change? That's a good question. So there have been some increases. We think about California, it's been drier and which causes more wind as well, but we've also seen, and this is only for wildfire, but some other perils as well, we've seen a migration towards higher risk areas. I think the number is like 70% of all new homes in California are built near the wildland urban interface. And so I call it the double whammy of wildfire in that you have more people in those areas that can cause fires to occur. And then once a fire occurs, there's more structures there to be damaged. And when I first started CoreLogic 12 years ago, the numbers were more like something like 50% of the fires are man-made, 50% of wildfires were natural. Well, now depending on who's doing the assessment, the number is more like 75 to 90% are man-made. And so we're seeing more people causing more fires. And again, there's more things to be damaged. So it's not only climate change. I'm not saying climate risk isn't an impact, but every bit amount of it is just there's more people at harm. Interesting. Well, thank you on that. We will end. And I would want to just remind everybody of the IMUA statement. And I want to say a big thank you to you for sharing your expertise, which is really important for our industry. It's crucial to all we're doing is understanding those models and using them to our best advantage. So thank you. And thank you to our attendees for wonderful, really smart questions too. Ali popped the survey link into the chat box. Hopefully you could take a minute to provide feedback. It's important to us. And let us know in particular with cat modeling, if you'd be interested in another topic within the area of cat modeling, because definitely Howard would be happy to cover those. Any suggestions, Howard? Again, I think we could dig into any one of these parallels a bit more and talk about it. There's also the use of models relative to... The word is slipping my tongue right now, the, gosh, anyway, the parametric model, parametric covers, that's becoming more and more popular as well in how these models are used on that. So there are definitely some opportunities for these models to be more... Better education can be done around these. Excellent. Well, and I'm seeing in the chat that the survey link's not working. I apologize. It will come out in the final email. Please do share your feedback. It's so important to us. We'll make sure that we send out the survey link if it's not working in the chat. And now before we end the CE credit, somebody asked, for SAMs or NAMs, we have CE credits for it, but not for any state license. So who, before you leave, who is the most last, best, and greatest? Who was the question that wowed you? Howard? That's on you. You're asking me? No, I'm asking you, who had the best question? Terah did. Excellent. Excellent. Well, congratulations to Karen. That's fantastic and well-deserved. You certainly were chatting in a lot of questions, including the one on AI, which I know is a very good topic, and you said that it was nice to be able to talk about that, which is, again, going to be very important with everything we're doing. So again, thank you to everyone. I think the chat link, like I say, will come out later. And it's in the chat now. So I'll leave it open, but Howard, again, thank you so much. Enjoy the rest of the day. We really appreciate you taking the time to share with us. Thank you.

Loss Control and Claims

Loss Control & Claims

Hurricane

catastrophe modeling

natural catastrophes

insurance industry

Howard Kunst

deterministic models

probabilistic models

forensic models

climate change

risk assessment

AI and machine learning

parametric covers