Premiums are rising. Insurers are leaving markets. But people keep building in risk-prone areas, and the climate disasters just keep coming.
Can insurance markets adapt?
In this episode, Shayle talks to Dr. Judd Boomhower, an assistant professor of economics at the University of California-San Diego and a faculty research fellow at the National Bureau of Economic Research. He studies how insurance markets are reacting to climate change. Shayle and Judd cover topics like:
- Why insurers are limiting coverage in California, Florida, and other high-risk markets
- How disaster insurance, unlike auto or health insurance, faces a flood of claims all at the same time
- How catastrophe models (or “cat models” for short) work and why AI and other improvements struggle the solve the fundamental problem: a lack of historical data needed to predict future events
- The challenges of private “black-box” catastrophe models that can’t be reviewed by third parties
- Reinsurance markets and why they’re not attracting more capital to shore up insurers
- The pros and cons of parametric insurance, an emerging category of insurance products
- Undercapitalized “fly-by-night” insurers that risk insolvency and failing to pay out claim
Recommended resources
- NBER: How Are Insurance Markets Adapting to Climate Change? Risk Classification and Pricing in the Market for Homeowners Insurance
- Brookings: “How is climate change impacting home insurance markets?”
Credits: Hosted by Shayle Kann. Produced and edited by Daniel Woldorff. Original music and engineering by Sean Marquand. Stephen Lacey is executive editor.
Catalyst is brought to you by Anza, a platform enabling solar and storage developers and buyers to save time, reduce risk, & increase profits in their equipment selection process. Anza gives clients access to pricing, technical, and risk data and tools that they’ve never had access to before. Learn more at go.anzarenewables.com/latitude.
Catalyst is brought to you by EnergyHub. EnergyHub helps utilities build next-generation virtual power plants that unlock reliable flexibility at every level of the grid. See how EnergyHub helps unlock the power of flexibility at scale, and deliver more value through cross-DER dispatch with their leading Edge DERMS platform, by visiting energyhub.com.
Transcript
Tag: Latitude Media: podcasts at the frontier of climate technology.
Shayle Kann: I’m Shayle Kann and this is Catalyst,
Judd Boomhower: Whatever statistical methods we want to throw at this thing. The basic limitation is that we’re trying to fill in gaps in the historical record. And so fundamentally these things are always going to be a little bit of a crystal ball
Shayle Kann: Coming up. How homeowners insurance is being affected by climate change. I’m Shale Khan. I lead the frontier strategy at Energy Impact Partners. Welcome. So as most of you know, I live in California, and if you live in California, as I’m sure is true, if you live in Florida or parts of the Southeast or maybe Texas or really a lot of places, then that homeowners insurance seems to be sort of unstable in a good chunk of the country. In the case of California, everyone that I know is maximum one degree removed from someone whose homeowner’s insurance was dropped, or at least had the price increased by some multiple because insurance companies recently reevaluated wildfire risk or maybe just made a decision to leave the state or something. Actually, that’s sort of the point. It seems to be fairly intuitive that climate change and the resulting increase in frequency and severity of weather events would make the insurance business tougher.
But I haven’t actually been clear on the mechanism what’s actually happening here. How are these companies adapting? Is it the models that are failing them and if so, why are they just making decisions with broad brushes, et cetera. So I wanted to talk through it because an area that I admittedly don’t know a whole lot about. So I found a great guest. My guest here is Dr. Juddson Boomhower. He’s an assistant professor of economics at UC-San Diego and a faculty research fellow at the National Bureau of Economic Research, and he’s been writing for a long time about the impacts of climate change on insurance markets. So here’s Jud Judd, welcome.
Judd Boomhower: Thank you. Great to be here.
Shayle Kann: Alright, insurance and climate change is the topic that I’ve been wanting to learn more about for a while, to be honest, but just haven’t gotten around to it. So you’re my foil. I think we’re going to focus mostly here on homeowner’s insurance in particular though I’m sure some of this applies more broadly. Let’s start with some history here. Are there in recent history, I guess, are there major events, climate events, weather events I guess, that have caused significant changes in how the insurance industry models or prices or covers homes?
Judd Boomhower: Yeah, absolutely. I mean I think that when you think about how climate change is affecting, especially people that live in the United States, one of the salient ways that we’re feeling it is through homeowner’s insurance, right? We have these increasing trends and disaster losses from all kinds of events, from floods, from hurricanes, from wildfires, from severe windstorms and hail. And we have seen really big increases in aggregate losses from those things, particularly in the last 10 or so years. And when you add all that up, it’s coming through the prices that homeowners are paying for insurance. So if you think about individual events that have really been important, we’ve had terrible wildfire seasons in California in 2017 and 2018, which are something that I’ve thought a lot about in my own research. We’ve had lots of bad hurricane events in Florida, we had Hurricane Ian, I think, and it’s also not so much any individual exceptional event, it’s the rate at which the exceptional events are coming, right? We’re having a lot of them. Things that used to be complete outliers are happening with greater and greater frequency, and those are adding up to really stress insurance markets.
Shayle Kann: So I mostly want to focus on where the industry is today or what’s coming. Obviously we’re recording this. What I dunno, is it two months since the LA wildfires, something like that. But since then there’ve also been a raft of other major weather incidents in other parts of the country. I guess the first question is, where is the state of the insurance industry right now when stuff like this happens? Is it economically really problematic for these companies? Are they prepared for it? I mean, we’re hearing all these stories of companies dropping coverage in locations, things like that, but are they hurting because of it?
Judd Boomhower: Well, I mean they’re hurting in the sense that their profits are not what they would like them to be. They’re certainly, they’re having a very bad time in terms of how their business is doing. I think kind of a subtext of what you’re asking is how likely is it that we’re actually going to see insolvencies or failures of major insurers as a result of these events? And for reasons that we’ll talk about, I think that’s less likely. In the short run, what is likely in the short run is that prices are going to need to go up returns to the equity investors and these companies that have equity investors are not going to be good. The insurers are not doing well in terms of profits and they are certainly choosing to pull back on the places that they’re willing to do business. So to the extent that they see areas where they have a big concentration of climate risk exposure and particularly where they think that they can’t charge prices that reflect that exposure, they’re definitely choosing to pull back on offering coverage. We’re not at the point yet where we expect major failures of property insurers.
Shayle Kann: When you talk about the pulling back coverage thing, this is one thing I’ve been wondering about. How sophisticated is that analysis? It seems like, I dunno, it feels fairly rudimentary in the sense that sometimes what you’ll see is one major event happens in one location, presumably that major event was predictable on some likelihood. And the result of that major event is that a bunch of insurers either use it as an excuse or just react quickly and say, okay, I’m pulling out of this market or out of this region or whatever it might be. Is that a rational choice? Is it a reflection of their poor analysis? How should we think about that?
Judd Boomhower: It’s a good question. When we think about how these firms are making these decisions, the basic thing to understand is that these risks are really hard to price and they put firms in a difficult position. So if you think about something like heart attack risk or auto accident risk, there’s a lot of historical actuarial data on those events that makes it pretty easy for insurers to know how much risk they’re taking on when they write a given policy. Natural disaster events are just fundamentally different because we thankfully have not had a thick enough historical record of disaster events that we can price these things using the statistical methods that are standard in other lines of insurance. And that means that we’re reliant on a set of tools that are much more kind of simulation and engineering based. And so chief among those is what we call catastrophe models, which are tools that insurers and consultants to insurers use to develop a best guess of the range of disaster outcomes that might affect a given customer in a given year.
And so when you say why are insurers pulling out of the places that they’re pulling out of, they have some sense of what their disaster exposure is that comes from these cat models. It’s not perfect. There is also absolutely a sense of reactiveness here that something bad happens and let’s get out of the place where the bad thing is happening. And that I agree that there’s this ad hoc, a little bit of an ad hoc sense to it, but fundamentally I think it comes from the fact that it’s a really difficult information environment. And so there’s other reasons we can talk about it, but that’s my high level take.
Shayle Kann: Let’s talk more about the cab models a little bit. Can you actually just go into a little bit more detail about exactly what are the cab models, how have they been built historically? And then this is one of those areas that you imagine two things you could imagine, one that whatever technique we use to build the cat models historically is insufficient for today’s environment because the climate change is causing more volatility in weather. And so maybe that breaks the cat models. I don’t know. But two, you could also imagine that we’ve developed a suite of modern tools, AI and so on that would make cat models better and there’s a lot of AI for weather forecasting and things like that. So I guess start with just how do the cat models actually work today and then is that a source of the problem or is it going to be the solution?
Judd Boomhower: Yeah, well, I mean the fundamental challenge that you’re trying to solve with a cat model is that if you look at a given house or a given city, you don’t have enough historical data to understand the likelihood of a disaster. If you have a disaster that occurs with a one quarter of 1% probability in a given year in order to really nail down the expected annual losses from that kind of event in a given place, you need hundreds and hundreds of years of data and we just don’t have that. And so the problem that we try to solve with cat models is we try to use the model to basically fill in those gaps in the observed data record. So if I’m trying to offer coverage in a given city, I have some sense that there’s some small probability that city’s going to flood or that city’s going to have a fire.
We don’t have the hundreds of years of observations for that city to know exactly what that probability is, but there’s another city in another state that looks the same. It has kind of the same vegetation, it’s kind of the same elevation, and maybe there’s a hundred more of those cities around the country. And if you start to put together the experiences of those different cities, then you can start to build a statistical model about what’s going to happen on average. But in order to do that, you’ve got to make assumptions about which cities are good proxies for other cities, which types of homes are good proxies for the dollar damages that other types of homes are going to experience when an event comes through. So basically the role of the cat model is to help you project out of sample to understand what the damages are going to be in a place where we’ve never actually seen this event, but we think there’s some possibility this event could happen. So in terms of how does the model actually do that, it’s going to take what’s called an event set, which is a real list of actual historical events that have happened, major hurricanes, major earthquakes, major floods, depending on what the thing is that you’re trying to model. It’s basically going to redraw from that event distribution randomly a bunch of different times to do this simulation based.
Imagine rerunning the world thousands and thousands of different times, and that kind of resampling from the historical event distribution is going to give you a shape of possible future scenarios and then you sort all of those possible future worlds that gives you a probability distribution of losses for a given property or a given city or a given country. So the last thing I would say shale to your question is the spirit of your question is what are we not getting right in cat models? And that’s a fair question and I do believe that there are opportunities on the engineering side to improve these things. Maybe AI gives us an ability to find trends in the data that we hadn’t been finding before, but it’s also important to recognize that we have kind of a fundamentally unsolvable problem here that whatever statistical methods we want to throw at this thing, the basic limitation is that we’re trying to fill in gaps in the gaps in the historical record. And so fundamentally these things are always going to be a little bit of a crystal ball.
Shayle Kann: Right. And I guess do we know have the cat models, maybe we don’t have enough data to answer this question, but have they been sufficiently accurate recently? Is there any question as to whether they’re good enough to, I mean I guess they are the best we have, but what do we know about their accuracy?
Judd Boomhower: They are the best we have. They’re the best we have, and I am beating up on cat models a little bit. I don’t want to come across as saying that they’re not useful or that we shouldn’t rely on them. They are the best we have. And so what we can do is when we have major events, major new events, we can compare the losses in that event to what the cat model said the losses were going to be. And I think that kind of post evaluation is useful. It’s limited because we’re observing one event and these are models, so there’s always going to be some prediction error for any given event. And it’s really hard to know if you see the cat model was off a little bit, is that what do we take from that one observation about the validity or accuracy of the overall model?
It’s still a useful exercise to do. The other challenge we have here though is that almost all of the models that are applied by insurers today are developed by for-profit private companies. It makes them all a little bit of a black box. It makes them all basically impossible for objective third parties to evaluate. And so we’re sort of dependent on these companies to tell us after the new hurricane or after the new wildfire, how accurate their model was. And if you’re cynical, you can imagine that there’s an incentive to at minimum be really public about how accurate the model was in the cases that it worked and maybe be a little less public about how it was in the cases where it was off. And these model are competing with each other as well. So there’s market pressures that make it a little hard to interpret what we hear about the validity of the models.
Shayle Kann: Right. And is there a big effort to improve the cat models? Obviously, like I said, there’s a ton of activity around AI for weather forecasting, which is a different thing from cat modeling, but have there been major advances or are there planned major advances?
Judd Boomhower: Yes, the models are getting better. There are a few major firms that have been doing this for a long time. They are making interesting investments to make the models better. I think the other thing that’s exciting is you’re seeing entry, so as the importance of this problem gets bigger, innovators are innovating and you’re seeing startups trying to come in and exactly motivated by what you’re saying, like, Hey, seems like this is really important risk. It seems like these models could be better. Is there something that I can figure out how to apply here that will give me a model that’s better, give me a big advantage, and I can license that to an insurance company or otherwise make a ton of money?
Shayle Kann: I’m curious how the cat models relate to pricing. So we’ve talked about insurers pulling out of markets. The other thing that’s happening is just the prices rising in some markets where all of a sudden there’s higher perceived risk. Do the cat models inform pricing in the sense that, I guess what I don’t have a clear view of, is cat model telling you the likelihood and severity of a given event, a hurricane or whatever it might be? Or is it also saying for this property at this address here is the likely damage and thus what the insurance risk would be?
Judd Boomhower: The model is delivering the latter. So what the model promises to deliver is for a given property in a typical year, what is the dollar amount of the expected losses that house or shopping mall or whatever it is, is going to suffer, and it’s even more ambitious than that. It’s also promising to deliver a probability distribution, so a variance of losses in a typical year and the co variance of losses for that property with other properties that you might be insuring in your portfolio, that co variance, that relationship between the probability of losses on one property and the probability of losses on another property is really important When we start to talk about reinsurance and total worst case losses. So the model is delivering a lot. We’ve asked these models to calculate a lot of parameters that are quite granular and that, so where does that come from?
Where does that dollar denominated loss prediction come from? The cat model is basically coming up with a probability that the house is going to face an event in a given year, which we sometimes call the hazard, and then the damages in dollars that the insurance company is going to have to pay if the house is affected by that event, which you could call vulnerability. And so one of the places that the models are really working on getting better in an interesting way is understanding how to price investments that people make in making their property less vulnerable. So if you live in the floodplain but you elevate your home, that’s going to greatly reduce the damages that you’re going to experience when a small flood comes through. If you live in wildfire country and you put a class a roof on your house or you manage vegetation around your home in a way that’s responsible, those things are going to reduce the likelihood that your house is going to burn down when a wildfire comes through. And so an ideal cat model would capture the effects of both of those things.
Shayle Kann: There’s another actor in this whole equation that I know plays a big role, and I honestly don’t fully understand how, which is the reinsurance market. Can you describe, I mean, what is reinsurance and then what is the role that it plays in pricing and coverage for homeowner’s insurance in major weather event driven areas?
Judd Boomhower: Yeah, absolutely, and this is very much at the core of the challenges that we’re seeing in these markets. Let me start by saying one of the fundamental differences between disaster insurance and insurance for other risks is that these are risks that are correlated, right? And so imagine being an insurer who’s writing insurance policies for everyone who lives in a given city. If you are covering health insurance, let’s say every person in that city has a 1% chance of having a heart attack in a given year, those heart attack risks are pretty close to independent, and that means the law of large numbers is going to be your friend. It’s extremely likely that in any given year you’ll pay claims for very close to 1% of the population, and it’s basically inconceivable that you would have a year where everyone in the city has a heart attack at the same time.
Disaster insurance is fundamentally different. If you think that you’re covering everyone in this city against a flood that has a 1% chance of happening in a given year but is going to flood the entire city, then 99% of the time you have no claims, which is great, but in one out of 100 years, the entire customer population is filing a claim. And so that huge variance of losses, that huge tail risk that you’re going to have to pay out a lot of claims at the same time is a fundamental difference in disaster insurance relative to other kinds of insurances. And that’s the problem that drives all of this discussion about reinsurance. Insurers need to make sure that they can access enough capital to pay these enormous potential claims with a high probability. We think about insurers trying to protect their solvency, and one way that they can do that is that they can hold a bunch of surplus capital within the firm, but that’s an expensive thing to do to basically put capital in low risk investments to be sure that they’re going to have that money when they need it. The other thing that they can do is that they can appeal to reinsurance markets, which are basically large globally diversified pools of capital that are able to hedge that risk of flood in that given city against tons and tons of other things that they’re invested in tons and tons of other potential losses to smooth away that right tail of the loss distribution and try and get back to the state where the law of large numbers is telling us that we’re not going to have huge right tail losses for the entity in a given year.
Shayle Kann: One thing that I’ve seen happen a little bit is in these markets where some of the larger insurers do pull out or they raise prices a lot, there’s an emergence of a new class of insurers, either those that are offering a different product like parametric insurance or something like that, or just new upstart, smaller ones. Is that a healthy dynamic? I mean, you want more coverage in the market, I suppose.
Judd Boomhower: Yeah, so we were just talking about reinsurance and one question that you might ask if you’re an insurance company is, well, reinsurance is expensive and difficult to buy. Why should I care if I as a company can’t pay my claims in a given year, then I’m going to go bankrupt and that’s terrible for my policy holders, but my losses are basically limited by the value of my company. That sort of coldhearted economic logic is absolutely something that’s been a problem in the insurance company for as long as insurance has existed. And so we do want to make sure that when companies are writing policies, that they actually have the ability to pay those claims in almost every possible realization of the loss distribution. So one of the things that we’ve seen in places like Florida is that there has been an increase in the number of really small insurers who are taking on property risk, and some of these insurers have capital balances that are potentially concerning.
And so there’s a whole weedy discussion of which credit rating agencies are reliable and not reliable. But the punchline of the story is that we’ve seen and increase in the market share in Florida of insurers that have questionable or potentially questionable ability to pay claims. And we have also seen as hurricane events have become worse, we’ve seen multiple insolvencies of these insurers. So we’ve actually seen these insurers go under not be able to pay all the claims associated with a given loss event, and that of course is breaking the promise of what insurance is supposed to do. These households paid claims, they expected that they’re going to be made whole when the disaster happens. And then that throws you into a whole different question of does the state step in and bail out those homeowners and do they take a haircut? But fundamentally, we want to avoid that whole problem by making sure that we have solvency regulation in place and good risk rating that’s making sure these firms are able to cover the losses they’re promising to cover.
Shayle Kann: I guess final question for you, I sort of mentioned it, but I’m curious, your perspective on the rise of stuff like parametric insurance, which as I understand it basically says, okay, if X happens, we pay you out Y and you don’t go through this complicated claims process, you can get paid out much quicker. There seems to be a rise in that, particularly as a result of climate driven events. How do you see that playing a role?
Judd Boomhower: Yeah, I think parametrics are really interesting. Parametrics solve a lot of the contracting problems that exist in any insurance market, be it climate or not climate. So whenever you have what’s called an indemnity insurance contract, that’s a traditional insurance contract that pays you whatever you lose, there’s a fundamental potential difference of opinion between the insurer and the customer in how much the actual value of the loss was, whether the loss was caused by the disaster or whether it was caused by something else, some lack of maintenance prior to the event, all kinds of stuff. And those fights lead to situations where homeowners feel like their insurer is paying them less than the true value of their claim, or insurers feel like the homeowner is trying to exaggerate the value of the claim. And those problems make people less excited to participate in these markets. And so the interesting thing about parametrics is that these are insurance policies that take away that whole question of trying to how much do we need to give you to make you whole after a disaster?
And it’s just a contract that says if some readily observable trigger happens, we will give you a pre-specified amount of money. And so if a hurricane with wind speeds above X strikes between this place and this place on the coast in 2025, we will pay you $20,000. That really simplifies a lot of the contracting side of the market. The challenge is it introduces what’s called basis risk. So what if a hurricane with wind speed x minus epsilon strikes, that’s probably still not great for me as a homeowner, but the parametric policy is going to give me no coverage. We didn’t actually hit the trigger. So the challenge, the needle you have to thread with a parametric policy is how do you make the basis risk small enough that it still provides enough insurance value to the customer without basically tiptoeing back into the world of something that looks like indemnity insurance.
Shayle Kann: Why is it that reinsurance is so hard to get and so expensive it seems like, I dunno, the financial markets would’ve solved for that somehow.
Judd Boomhower: That is the million dollar research question right now and the million dollar policy question. So in your cartoon model of finance or economics, this is a problem that’s not that hard to solve because there are a lot of risks in the world and the beta on climate risk is pretty nice relative to those other things, it’s pretty uncorrelated. So why aren’t there more entities that are waiting in and selling reinsurance and pushing the price of this product down? We really don’t understand the answer to that question, and it’s something that we absolutely have to figure out. So if you look at people that study reinsurance, the persistent fact is there’s not a lot of capital entering these markets. And is that because there’s not enough people that feel confident that they understand the risks? Is that because there’s something that looks like a cartel between a small number of players who are benefiting from market power and keeping prices high? Is that some other reason? Whatever it is that’s creating friction in these reinsurance markets is in my view, one of the absolutely most important things that we can figure out to make property insurance work because that inability to lay off that tail risk to the global capital markets is really, really at the core of the problem that we’re seeing in property insurance.
Shayle Kann: Alright, Judd, this is super interesting, more to talk about in Insurance world, I’m sure. So we’ll have another opportunity, but thanks so much for the time.
Judd Boomhower: Thanks, Shayle. This was great.
Shayle Kann: Dr. Judd Boomhower is an assistant professor of economics at UC-San Diego, and a faculty research fellow at the National Bureau of Economic Research. This show is a production of Latitude Media. Head over to latitude media.com for links to today’s topics. Latitude is supported by Prelude Ventures, prelude backs, visionaries, accelerating climate innovation that will reshape the global economy for the betterment of people and planet. Learn more@preludeventures.com. This episode is produced by Daniel Woldorff Mixing and theme song by Sean Marquand. Stephen Lacey is our executive editor. I’m Shayle Kann, and this is Catalyst.
