Critical Point Episode 29: Understanding the impact of flooding on mortgage risk

Announcer: This podcast is intended solely for educational purposes and presents information of a general nature. It is not intended to guide or determine any specific individual situation and persons should consult qualified professionals before taking specific action. The views expressed in this podcast are those of the speakers and not those of Milliman.

Rebecca Driskill: Hello and welcome to Critical Point. My name is Rebecca Driskill, and I’ll be your host today. Milliman has just hosted it’s first global Climate Resilience Forum, looking at the ways we can anticipate, measure, and communicate climate risk. In that vein, my guests today have recently written an in-depth paper for the Society of Actuaries that looks at flood risk and sea level rise and its potential impact on the mortgage market.

Rebecca Driskill Dave Evans and Cody Webb are both consulting actuaries with Milliman, specializing in flood risk. Good morning, guys.

Dave Evans: Good morning, Rebecca.

Cody Webb: Good morning. Thanks for having us.

Rebecca Driskill: And Andrew Netter is a senior financial consultant with Milliman's mortgage group. Morning, Andrew.

Andrew Netter: Hey, Rebecca. Hey, everybody. Thanks for having me.

Rebecca Driskill: And Brandon Katz is executive vice president of catastrophe modeling firm KatRisk. Hey, Brandon.

Brandon Katz: Hi, great to be here.

Rebecca Driskill: Flooding stands out among other natural catastrophes in terms of the risk it poses to the financial health of the average U.S. homeowner. And the research you all did, first it examines residential exposure to flood risk countrywide, it also explores the relationship between flood insurance and mortgage risk, and then also looks at how that risk can impact financial instruments designed to cover mortgage default exposure and those portfolios. So we’re going to try to touch on all that today, but first I’d like to start with a high-level overview about how flood and mortgage risk intersect. Andrew, can you talk a little about that?

Andrew Netter: Sure, I can start with that Rebecca. So, taking a step back, Milliman's mortgage practice has been around for over twenty years and a large part of what we do is evaluate mortgage credit risk. Generally, you can think about that as estimating the likelihood that a mortgage or a group of mortgages defaults and fails to pay back the full amount of money lent. Historically, the industry standard of modeling that type of risk has been to focus on borrower credit and financial strength. So, things like credit score, debt-to-income ratios, as well as macro economics where borrowers may be unemployed and not able to pay back their mortgage. Where flood risk and the impacts of climate change becomes more important is the fact that those things could also be a catalyst for a borrower not paying back their mortgage. To the extent that there's uninsured losses, a borrower may have to use their own personal reserves to fix the property. Those things can spill over into mortgage credit risk and cause a loss that an investor or a stakeholder otherwise wouldn't be contemplating.

Brandon Katz: I think it's also interesting to think about what the history is, you know, kind of going back on why flood and mortgages are even linked in the first place, because, let's say, by and large, when you're thinking about buying a mortgage, you know, "Is my house going to be impacted by some sort of potential natural catastrophe?" Is that really something that really enters into people's conception. So, it's useful to ask the question, "Why are flood and mortgage even linked in the first place?"

I think a lot of that goes back to the 1970s when the National Flood Insurance Program was even initiated. And, at that time, in the 1970s and the 1980s, that's when we started having a requirement that any federally backed mortgage had to have flood insurance associated with it, typically for the National Flood Insurance Program. And, up until about ten years ago, that was really the vast majority of the flood insurance and the flood policies that were in the United States were through the National Flood Insurance Program, mainly because they were the only ones that were really doing wide-scale modeling in the United States. So, they typically kind of tranched everything into one-in-one-hundred-year flood zones and then things that were exceeding one-in-a-hundred-year flood zones, and required you to have a policy if you were within a one-in-one-hundred-year flood zone. So, I think that's an interesting thing to start with, because that's the whole reason why we even link mortgages with floods now, is because of some of those initial federal requirements.

Rebecca Driskill: And, so, Dave, Cody, your practice has been modeling flood. Now, what are you seeing change and can you talk a little bit about the work you guys have been doing?

Dave Evans: I think, in general, there's just been a lot more activity with flood, but one thing, in particular, going back to what Brandon said, you had the National Flood Insurance Program since 1970, doing their modeling, you didn't have a lot of other flood modeling. And, in the last five to ten years, there's been a lot of advancements and people are starting to change the narrative as far as the perception that if you're not in a flood zone, then you don't have flood risk. And then that naturally transitions into mortgages, where the requirement is that if you're in a high-risk flood zone, if you're in the special flood hazard area with a federally-backed mortgage, then you need one. But, now, everybody's realizing when you cross that magic boundary outside that, you're not suddenly immune to floods. And, so, that starts to incite this question of "Okay, what is the risk to mortgages and to lenders and to everybody in that ecosystem?"

Rebecca Driskill: Yeah, I mean, that sort of-- that brings up a really interesting point in this idea of quantifying the climate risk here. Can you guys all talk a little bit about what was the goal of the research?

Dave Evans: Yeah, I'd say-- I think we had a couple goals, but I think if you want to look at the one directly related to flood and mortgages, you know, the idea is that there's really not a lot of public information out there or anecdotes or anything about how people are quantifying the flood risk to mortgages. And I think there's a good reason for that. It's a topic that's just kind of emerging. It's difficult. And, if you look at the past, the risk has not really materialized very much in a significant way. So, you're predicting something in the future, a risk that has never really materialized too much in the past. So, that makes it really challenging.

And, so, what we wanted to do is we knew that Andrew and his colleagues had a lot of experience predicting mortgage risk and we work a lot on the property side. And Brandon's firm is very good at what they do in modeling the hazard of flooding. So, we thought that with all these pieces, the tools are there to explore this future and explore it in a more sophisticated manner. And, so, we felt like we took a good step there with this paper of showing how you can couple your-- the growing flood risk measurement that's out there and that technology with sophisticated modeling of the mortgage risk and come up with a better picture of what the actual risk is to mortgages and to the people that hold those mortgages in their portfolios.

Andrew Netter: I think what Dave highlighted is that it was bringing together two pieces of the puzzle to create that framework to add flood and climate risk to mortgage analytics. Oftentimes, you know, the mortgage group is separate from the cat modeling group. And those things don't speak to each other. Here, the paper, one of the strengths is that it does give that framework so that others, you know, can follow that and have a more holistic view of mortgage risk, both credit risk as well as the climate change and flood impact. In general, what our paper found is that there is risk there // a lot of times that's concentrated in specific geographic regions. But, overall, it's not something that you can just overlook. And, as climate change has stronger and stronger impacts, it does need to be part of a holistic approach to mortgage modeling especially to the extent that somebody has a portfolio concentrated in an at-risk region, whether it be for floods or other perils.

Rebecca Driskill: What surprised you as you wrote the paper, as you got the results of your research?

Dave Evans: One thing that surprised me was Like, eighty-seven percent of losses are not covered by the NFIP and the private market's really not covering that very much at the moment or bridging that gap at the moment. So, I think that was surprising. And then that kind of falls itself into when we looked at, you know, the risk to individual mortgages, if you looked at how large flood losses are and how much those could impact the homeowners risk of default if a flood loss was similar to other non-catastrophic impacts to their property value, it's very large at the individual level. So, flood is largely uninsured and it's very disruptive at least when you look at it in the context of how homeowners have reacted to changes in their property value. The magnitude of that was surprising.

Brandon Katz: Oh, I was just going to add on top of that. You know, I think the magnitude was very interesting and one of the things we think about when we think about the catastrophe models is both frequency and magnitude. So, you know, looking at some of the statistics that came out of the paper, I just jotted a few of these down, so, looking at the New York area, for instance. We modeled two primary, different sea-level rise scenarios. Sort of a reasonable medium size and a reasonable sort of on the higher range. And, you know, the New York metro area, for instance, with the medium-type scenario, we had an increase of around thirty percent in losses. For the high scenarios, almost a doubling with losses.

So, the other thing we looked at was a one-divided-by-five-hundred every-year probability of loss. So, that's just, basically, every year there's a probability-- some probability that this kind of event can occur.. And, so, why does that matter? if you have more events, smaller events, that are occurring more frequently that are actually causing a higher loss? Every year, you're going to be seeing more hurricane loss. So, it might be small losses, but every single year you're going to be seeing more and more hurricane losses. So, I think that if you look at this from that standpoint, if people paying attention for potential legislation, looking at those lower term-period losses are the ones that you actually want to see.

But, yeah, essentially, what we're looking at then is more frequent, lower term-period storm surge events as we see sea level rise increase in places, like, New York, and other places up and down the eastern coast.

 

Rebecca Driskill: That's really interesting, because you kind of see what's happening from a consumer perspective in the news in terms of Hurricane Laura and the Gulf Coast, but sort of being able to zoom out and understand what are the trends and then think about the larger picture of, the paper found that building losses are expected to cost more than $7B annually, let's think about how much of that is insured. I don't know, Cody, can you talk a little bit about who bears the risk?

Cody Webb: Well, I think, historically, for uninsured losses the homeowners bear a significant fraction of the risk. I would say there's been a lot of state and federal assistance money that has gone to partially make homeowners whole, who have had these events. It really varies event by event, whether these types of grants for people who are uninsured to re-build will be awarded. So, to a certain extent, the gap, the insurance gap has been filled by assistance dollars, state and federal assistance dollars. But there's no clear indication for the risk going forward of whether policy holders would expect to receive those funds in the future.

Rebecca Driskill: So, we've been talking a lot about flood, but I'm also interested to talk about wildfire given what's happening in California and Oregon and the whole West Coast. Can you talk a little bit about how wildfire and mortgage could intersect and what you're seeing right now and your thoughts on that?

Cody Webb: As Brandon was mentioning earlier, one interesting thing about flood insurance is just sort of the history of flood insurance and the NFIP and the events that caused flood insurance to be sold as a separate policy from homeowners insurance. So, all over the country, generally speaking, homeowners insurance is sold and it excludes flood. And that has certain pros and cons. So, one of the benefits of that is that people in these flood-prone areas can get homeowners insurance on the rest of their property, because the flood risk has been decoupled from the homeowners policy and all the considerable challenges associated with quantifying and pricing flood risk are not presented to the insurance company.

In California, earthquake insurance and wildfire insurance are kind of two separate perils that affect California that parallel the flood situation to an extent where, like flood insurance in the United States, the earthquake policy in California has been decoupled from the homeowners policy and a lot of people lack earthquake insurance, but they're able to get homeowners insurance in those areas. However, wildfire insurance is a different situation and wildfire insurance is included with the policy. And some of the cons associated with that is as the risk goes up, because this is part of the standard homeowners insurance policy, homeowners insurance availability issues have started to crop up as a result of the wildfire and as a result of insurance companies pulling back from the state.

So, now you have a situation where a large number of people in California are unable, or may become unable, to obtain homeowners insurance at all. And, of course, this is just crucial to the-- insurance is crucial to the mortgage market, because insurance secures the collateral underlying the mortgage. And if we end up with a situation where a lot of people don't have insurance for whatever peril, for whatever reason, then we can end up with a parallel situation to this flood thing, where there's substantial mortgage risk resulting from a lack of insurance that people are able to obtain.

Brandon Katz: Yeah, and I think it's also interesting to t-- of course, I'm thinking more on the catastrophe modeling side, you know, one of the ways that we actually really understand should something be insurable and should be something be covered, you know, one of the things we think about there is what has happened in the past? And that's, typically, how we have done things, like, decide, "Is one region too risky for flood or too risky for earthquake or severe effect of storm? Should there be exclusions on a policy? Should a mortgage require this or that?"

With wildfire in California, and up along the West Coast as well, you know, there's a couple things to think about with climate change, because it comes up a lot. You know, are we having more drought days? But, also, really integral is have people begun to move into places that are closer to that-- what's called the WUI, wild-urban interface. You know, are you getting too close to the wooded areas? And that's one thing that's actually been really difficult to quantify, because you can't really use the fires of the past to predict the sort of losses that you have in the future, because people are living in new areas. And that's actually one of the reasons why catastrophe models are so important, because you're able to simulate potential past and potential future events on any exposure that you're able to put in there. So, you can actually, then, place those buildings in locations where they didn't exist and run, say, some fire of the past using the exposure of today. So, it's very important to be able to model these things and not just look at what's happened in the past. Because if you just look at what's happened in the past, you're going to be underestimating the potential loss because people simply live in places they didn't use to live. And that's why a lot of modelers, KatRisk included, we're working on a wildfire model, a lot of other modelers are working on their wildfire models. Some are out, some are not out. But that's why it's really important to not just look at what's happened in the past, but to be able to model what could potentially be happening in the future. And climate change isn't just more drought days, it's also people are living in different places that, maybe, are more risky than it used to be.

Rebecca Driskill: What can we learn or what should we be trying to understand about mortgage risk and the impact of these perils, like wildfire and flood, and how that's been changing?

Dave Evans: I think one interesting thing we found is that if you look at what's covered in a basic homeowners policy today, some of the-- what we would call "the major perils" like fire or wind and hail, those are-- all have a similar magnitude of expected losses as flood losses, yet flood is not covered in the policy. So, I definitely think that's an interesting thing to think about is that all homeowners are required to purchase coverage to cover a fire that might happen, whether it's catastrophic or non-catastrophic, One key difference between them is just that floods are a lot more widespread. So, no matter where you live, you could be at risk of flood whereas wildfire tends to be much more concentrated, right, in the West, obviously, and you're not as worried about a hurricane if you live farther away from the coast. But flood is really something that is spread out across the entire country. And that's part of the reason, too, with our study, why we wanted to-- when we first talked to Brandon and KatRisk that we wanted to put something out there that quantified some of the flood risk across the entire country. You see studies that talk about climate change and the impacts to Florida for flooding or the possible risk of flood losses in the Gulf. And those are all very important, but I think it's also important to get more information out there about just the magnitude of flood risk in the whole country, because there really is flood risk in every county of the United States.

Rebecca Driskill: What are the main takeaways in terms of potential actions or potential to just sort of things to be aware of and look out for and start to try to anticipate and communicate to the public?

Brandon Katz: Yeah. I think one of the main points is that the past does not dictate the future. You know, looking at this study, obviously, we had Super storm Sandy that impacted the New York area, but the study very clearly shows that places like in the New York-Jersey City sort of area that the risk there is actually going to be much higher and we're going to be having much more frequent losses, as we alluded to earlier. So, I think that, you know, David very clearly mentioned, I think correctly mentioned, that many studies focus on Florida and what's going to happen in Florida. But it's really not just Florida. So, there are areas of New York that we should be thinking about raising sea walls in order to mitigate the disaster. Super storm Sandy could give us this sort of preview of what that could potentially look like. That was not a huge hurricane, but it was a very, very larger term period hurricane when it comes to inland flood losses, which we don't always think about when we think about hurricanes. A lot of times we think about storm surge, which is, you know, in general, what this study focuses on, but there are other hazards, like inland flood, just pure precipitation-induced flooding that we should be thinking about.

But, yeah, I definitely think that, you know, there is a lot of storm surge risk due to sea level rise. It happens in parts of the country that we don't typically think about and that's what this study highlights. It tells where we should be looking and where it might be a good idea to be investing some dollars to make sure that maybe we can try and mitigate some of this risk. And it also tells us that in the future, and this is just thirty years out-- this isn't that far out-- that there are some portfolios of mortgage risk that could become very dangerous to be holding onto and we should be thinking about regulation to make sure that we don't have any sort of collapse in terms of our mortgage-backed securities. We don't want to have any major issues or any surprises, and that there are methods and models for us to be able to quantify and understand and prepare for that risk that could be coming in our potential future, whether it's in Florida, whether it's in Maine, whether it's in North Carolina.

Dave Evans: Yeah, I think the last thing you said, Brandon, is-- a kind of key takeaway for me is that there's always-- I think we can do better as far as measurement, "we" being kind of everybody, what's out there collectively. And it's a difficult problem, especially when the past really isn't predictive of the future, or, at least, we don't think it is. But that there are ways to try to understand it and be more rigorous about that. So, I think that's an important takeaway. And, then, I think the other part is-- and, you know, this is something that gets heard a lot, but I don't think it can really be said enough is just that the statistics of the amount of people that are not insured for flood is just-- I don't know if I'd call it "shocking" anymore, but it's very large. We have a large protection gap in this country. You've got people being insured for all kinds of things and they still-- one thing we talked about in our study is that people probably think they have homeowners coverage, and there are other surveys that indicate this, when they don't-- I mean, that they have flood insurance coverage in their homeowners. So, I think we just have a lot of problems with our system that are-- whether it's, you know, private or public or regulation, it's not optimal when ninety percent of the time people's flood losses are not covered by insurance or at least the risk going forward. And with climate change and with sea level rise and other impacts, it's not going to go away any time soon.

Rebecca Driskill: So, we've been talking about flood risk as a peril and I'm interested, Brandon, from a catastrophe modeling point of view, what's changed? Why are we sort of learning more about private flood insurance and flood risk modeling now, when ten, twenty, thirty years ago we weren't really talking about this at all?

Brandon Katz: Yeah, I remember, you know, at my first job when I first started doing flood modeling, I kind of asked the question, you know, "Why are we developing flood models for other countries? Why not the U.S.?" It was a question I was certainly curious about when I'd first gotten out of university. And the big reason-- you know, there were sort of two: There was one, "Well, the NFIP is already doing it, so why should private insurance bother?" And they were sort of subsidizing kind of everything. And number two, it's very, very difficult to model these very high resolution hazards. So, when you're modeling something like wind or hail or even wildfire, as you can imagine, it's a relatively course, course hazard. And when I say "course hazard," I just mean that the hazard that I'm feeling at my house is the same as the hazard that's being felt several houses down on the block. Wind, for instance, is modeled at, say, you know, hundreds of meters or thousands of feet across. So, the hazard that I'm feeling is the same as the hazard that's being felt at the next block.

Flood is very different. Flood's a very high resolution, very granular risk. So, where I'm sitting at my desk right now, versus the street in front of my house, I could be dry in my house, whereas the road could be under one or two or three feet of water. So, we have to model flood at a very, very high resolution. You can't do it at a course resolution. And what this meant is that it was very difficult to model from a computation standpoint. Computers were not strong enough. So, as an example, for us to model the flash flood component of flood risk, typically called the "fluvial risk" we actually had to license time on the Titan supercomputer, which at the time was the fastest supercomputer in the world. And we had to run our model on there for several months on hundred of GPUs, which is a new way of doing computation. And that simply wasn't possible even when I first started working in catastrophe modeling, you know, roughly ten years ago. So, the rise of fast supercomputers is the only reason why we're actually able to quantify the risk, as opposed to kind of using the piecemeal studies that the NFIP did in the past, which are very, very good, but some of them are very dated. Some of them haven't been updated since the 1970s and 1980s. And, so, I think it's very important to say that flood is a quantifiable risk now. It's something we can actually look at and give some numbers and some differentiation, whereas even ten years ago that was a very difficult proposition.

Cody Webb: And I would just piggyback on that by saying policymakers should be aware that the current policies in terms of flood insurance quantification and requirements don't reflect these advances in computing and science. So, we look at a situation where we say, "How can so many people with significant flood risk not have insurance?" And it's because the policy was created at a time before all this modeling and quantification, that Brandon mentioned, was possible. So, policymakers should consider kind of updating the policies to utilize the latest science in terms of the risk quantification.

Rebecca Driskill: We're dealing with wildfires in the West, there's flooding at a national level. What is the next kind of research or quantification that you think would be helpful? Like, what would you guys really want to be working on right now to really help drive this conversation forward?

Brandon Katz: This study really looks at storm surge and how is it affected by climate change, specifically sea surface rise?

Brandon Katz: So, the next logical step would be now we're not looking just at the East Coast, at how storm surge is being impacted by climate change, but now we should also be looking at how are inland flood and flood policies being impacted by climate change? We've done a little bit of work there looking at, basically, as temperatures increase, the holding capacity of clouds and air to hold water vapor increases, which can increase actual rainfall from these types of events. So, looking at global climate change and global temperature change. So, you know, as sort of the back-of-the-envelop calculation, we actually ran our models by increasing some of the flood impacts for these different climate change scenarios. These are the same scenarios we actually used in the study and, for example, in Louisiana in the Year 2050, you know, we can see between a fifteen to a twenty-five percent increase in average annual loss. Looking at the Year 2100-- so, going even further afield-- there can be a thirty-five to forty-five percent increase in average annual loss. You know, Texas is a bit less than that. As an example, five to fifteen percent in 2050 and it going up to around twenty to thirty-five percent in the Year 2100. So, you know, kind of in quick summation, I think that inland flood is the next thing to look at and we're already seeing some very strong signals that show that the losses can be increasing in different regions of the country. And we're going to be seeing a very defined signal, climate change signal, in inland flood losses that will be affecting policies and mortgages in the future.

Andrew Netter: I'm looking forward to closer integration between catastrophe modeling and different perils and mortgage credit risk analysis. You know, historically, it's been very borrower-centric. We've seen over the years and through this analysis that there's real risks other than just borrowers losing their employment and not paying their mortgage. So, closer alignment of these two modeling disciplines can really just yield benefits for the industry as a whole in understanding the risks that underlie these financial instruments and securities.

Cody Webb: I would just say, you know, I think it's becoming clear the financial consequences of climate change are going to be very large and I would like to see things go from a place of just quantifying the future impacts to doing more cost-benefit analysis for society in terms of kind of determining what the real costs and benefits of the various things that can be done are. So, and I think the first step is, as we're doing, kind of quantifying the potential risk in the future, and that'll give us a good foothold and a good opportunity to consider the quantitative impact of mitigation strategies and insure that money is well spent.

Rebecca Driskill: Well, Andrew, Cody, Brandon, Dave, thanks so much for joining me today. You've been listening to Critical Point, presented by Milliman. To listen to other episodes of our podcast, visit us at Milliman.com or you can find us on iTunes, Google Play, Spotify, or Stitcher. See you next time.