Making toxic pockets of residential real estate more visible

An important new paper shows where and why rapid repricing is more likely to happen

Yesterday, someone called me up to talk about the profound water-related risks to Charleston, SC, where it flooded 75 times last year compared to 5 or 8 times in 1989-1990. People in Charleston understand that sea level rise is real and accelerating, because they see the effects downtown so frequently. Even as much of the nation broils this week, though, there's still a good deal of looking away from the physical risks coming toward us.

We change our minds and take action only when we see something for ourselves. And that's why Prof. Ben Keys's recent co-authored paper, "Property Insurance and Disaster Risk: New Evidence from Mortgage Escrow Data," is important.

By doing some clever backflips and data scrounging, he's able to show in a detailed way that people in riskier zip codes are paying more for home insurance. He can show that these amplified costs have climbed particularly quickly since 2020. And he's able to demonstrate that much of this price hiking is essentially a pass-through of what it costs insurance companies to share their risks with global reinsurance companies. If you're in a risky zip code in Tornado Alley, in Oklahoma or North Texas, or along the East Coast, your home insurance costs are climbing.

Individual homeowners see their yearly rates going up, of course, but the property insurance market as a whole is frustratingly opaque. There is no grand database showing which company charges what price for what form of coverage. It's all proprietary, fractured, and invisible, as I've written in the past. The companies selling these products have insights drawn from climate models—and the global companies that insure insurance companies, the reinsurers, essentially have the freedom to raise prices when they see risks: US property catastrophe reinsurance prices doubled between 2018 and 2023. 

None of this is necessarily malign. It makes sense that financial sectors that have these insights and have the power to reprice their products will do so when they can. The insurance and reinsurance businesses are on the front line of repricing climate risks. 

Think of this paper as yet another warning sign. It's now easier to see where rising insurance costs are likely to play a role in triggering a host of interconnected consequences: lowering property asset values in hard-to-insure pockets, undermining legions of property tax-dependent regimes, weakening confidence in municipal bonds, shifting increasing risks to undercapitalized state insurer-of-last-resort entities, shifting increasing risks to Fannie Mae and Freddie Mac, and potentially infecting many elements of our thickly interconnected financial system.

From the outside, Keys's paper is slightly funny. He's saying, "Okay, insurance companies, you won't tell us researchers what you charge for what, so we're going to reverse-engineer and infer as much as we can from millions of records of escrow payments!" Instead of looking head-on at actual information about particular coverage from particular entities (which isn’t available), Keys and his co-authors got access from CoreLogic to data about millions of monthly payments homeowners made between 2014 and 2023 into escrow accounts for almost 20 million mortgaged single-family homes. Usually, those payments go to lenders or loan servicers, and are for a lump sum each month. That lump sum covers principal (amount paying down the loan itself), interest, property taxes, and insurance premiums. 

By using data about millions of individual loans, Keys was able to subtract the first three elements from the lump sum—leaving him with an amount he could infer was being paid for property insurance. (Keys had to do some regression magic based on what he knows about VA loans to figure out which of these homeowners might also have to pay for mortgage insurance, and he subtracted that amount as best he could.) Then Keys was able to control for many other facts that might have affected insurance premium costs: property value, loan amount, borrower characteristics. He was also able to create a metric that tracked changes in insurance premiums for the same loans over time. All of this allowed him to create a dataset with 48 million observations of 12 million loans that allowed him to isolate the impact of location (zip code) and time on insurance premiums. 

Then, after accounting for inflation, he overlaid all those observations with zip-code level risk data from the First Street Foundation about wind-related damages from hurricanes and damages from wildfires, plus some other disaster risk data from the National Risk Index. He added in zipcode-level demographic data, removed outliers (very high or low values), and collapsed all this monthly data into quarterly data points.

All of this manipulation and inference allows Keys and his co-author to say that they know something about the relationship between disaster risk and insurance costs for a significant portion of single-family homes in different zip codes in the US, while controlling for (eliminating the influence of) house prices and other factors. 

Here's what he finds: 

• Disaster risk has a significant impact on premiums. Insurers apparently are directly pricing in climate risk, and insurance costs seem to vary in zip-code-level ways that mirror First Street's risk perceptions.

• Even after considering state-level differences in regulatory structures or number of natural disasters, disaster risk still has a significant effect on premium costs.

• The relationship between disaster risk and insurance premiums is getting stronger over time, especially in high-risk areas. Disaster risk became a more important determinant of premiums after 2019. In essence, the price tag for disaster risk went up by 2/3rds between 2018 and 2023, meaning that living in a high-risk area can lead to a substantially higher insurance cost compared to previous years. 

Here's a picture of the entire country's results. Notice Tornado Alley. Notice Florida and the Gulf Coast. Notice the East Coast:

• More than a third of the total premium increase in the riskiest zip codes between 2018 and 2023 can be attributed to higher reinsurance costs.

• The pass-through of reinsurance costs is extremely concentrated in disaster-prone areas in states with more reinsurance exposure (like Florida)

  

• Zip codes with higher median incomes tend to have higher average premiums, but the income burden of higher premiums is greater in places with lower median incomes:

  

• Zip codes with a lower percentage of white residents tend to have slightly higher average premiums.

• The top 5% of riskiest zip codes can expect annual premium increases of over $700 per year due to higher disaster risks.

If anything, this research is likely underestimating the likely effect of disaster risk on costs to homeowners.

Keys can't say anything about how high deductibles are, how narrow the coverage is, or how likely to be capable of paying claims a particular insurer is itself. This data reports only a total imputed quarterly amount paid for home insurance. Homeowners in high-risk areas might choose higher deductibles to lower their premiums, but then face even higher out-of-pocket costs in case of a disaster. They might reduce their coverage to the minimum to make their premiums more manageable, leaving them exposed in case of a disaster. They might sign up with a fly-by-night insurer just to keep their loan in place and allow their servicer to sell the loan on to Fannie Mae. 

At any rate, this research should send a detailed, visible signal that focusing on adaptation and relocation—actually lowering risks in areas we already are being told are risky—should be a central priority. The cost of climate risks is already being felt by homeowners who hold mortgages in risky areas. At the very least, we need to stop increasing these risks by allowing continued building in risky places. 

And surely we should have better public data about individual zip code insurance costs and risks to homeowners. The repricing, when it comes, will have brutal effects on many Americans.

If only we could plan ahead.