The price of your wildfire or hurricane coverage often has less to do with your particular house than you might expect. Much of it comes from your address being fed into a catastrophe model, software that simulates thousands of possible disasters.

That isn't a scandal. It's how hurricane insurance has been priced for a long time, and the approach is now moving into wildfire. But it changes what a premium is. Part of the premium is the output of a model, and models get revised.

What a catastrophe model does

A catastrophe model is a large simulation. It generates thousands of possible storms, fires or earthquakes, most of which will never happen, and runs each one against a set of buildings to estimate how often a given property is likely to be damaged and how badly. It has three working parts. A hazard piece estimates how often events happen and how strong they are. A vulnerability piece estimates how a particular kind of building fails under wind, embers or shaking. A financial piece estimates what the policy actually pays once deductibles and limits are applied.

Insurers use the results in deciding what to charge, where to write policies, how much capital to hold and how much reinsurance to buy. Many insurers license their models from a small number of outside developers rather than building their own. So when one developer revises its model, the view of risk can shift for a large share of the market at the same time.

When the software changes

In June 2023, Moody's RMS released version 23 of its North Atlantic hurricane models. Florida's hurricane-model commission certified it that same month. The update revised event rates across the whole Atlantic basin. That covered both the long-term rates and the medium-term rates, which fold in recent and near-term climate conditions and which Moody's treats as its default view. The update also changed how buildings are expected to respond to wind.

Guy Carpenter, a reinsurance broker, ran the new version against a sample of its US clients' portfolios and published the results in July 2024. On frequency alone, the changes were modest, about plus or minus 2–3% under long-term rates, with an increase in the Gulf and a small reduction in Florida. The bigger driver was vulnerability. Several building types, including multifamily buildings and structures built between 1995 and 2020, came out more prone to damage. Modeled residential losses in the Mid-Atlantic and Northeast could moderate or fall. Guy Carpenter's conclusion was that carriers "will need to evaluate how v23 affects their own view of risk," with consequences for underwriting, risk management and how much risk they transfer to others.

To be fair about what changed: a model update does take in new history. Version 23's medium-term rates added the 2021 and 2022 seasons, including Hurricanes Ida and Ian. What stands out to me is that the update re-estimated risk across the whole basin at once, including for the many buildings those storms never touched.

Guy Carpenter added an important caveat. It did not expect version 23 to shift catastrophe reinsurance pricing significantly, because many risk-takers had already adjusted their baseline loss estimates. A new model is not automatically a new price.

California writes the model into the rules

California held out for decades. For about 30 years, state rules had insurers base the catastrophe portion of wildfire rates on historical losses, according to the California Department of Insurance. On December 13, 2024, the department posted a final regulation allowing insurers to use catastrophe models in wildfire ratemaking. The condition is that major insurers increase their comprehensive policies in wildfire-distressed areas to at least 85% of their statewide market share. Consumer Watchdog, a consumer advocacy group, has argued that loopholes in the rule let insurers raise rates without writing meaningfully more coverage, and it has called the models "black-box."

A model has to pass the department's review before an insurer can use it. Verisk's wildfire model completed review in July 2025 and Moody's in early August. By August 2025, the department said it had finished reviewing models from Verisk, Karen Clark & Company and Moody's, which made all three available to every insurer in the state.

My read is that this is the clearest sign of the shift. A regulator that required insurers to price wildfire by looking backward for three decades now accepts a forecast, as long as it has inspected the model first.

The same idea, aimed at buyers

Models now reach homeowners directly as well. First Street publishes property-level risk scores from 1 to 10 for flood, fire, wind, heat and air quality, and frames several of them over the next 30 years. Zillow began showing those scores on for-sale listings in September 2024. It removed them in late 2025 after the California Regional Multiple Listing Service questioned the accuracy of First Street's flood projections. Listings now link to First Street's own site instead. The listing service also asked Realtor.com, Redfin and Homes.com, which had shown similar First Street data, to remove predictive flood numbers.

Since then, the dispute has moved into court. According to Insurance Journal, Nebraska sued First Street in August 2026, and Florida's attorney general followed on October 23, 2026, alleging that its scores rest on flawed science. First Street's CEO has said its models are built on peer-reviewed science and that the full methodologies are published. Neither case has been decided.

Who gets to look inside

Better models are not a bad thing. Pricing risk accurately is the point of insurance, and a more precise model can cut a premium as easily as it can raise one. In Guy Carpenter's analysis, version 23 could lower modeled residential losses in parts of the Northeast.

Still, it puts a lot of weight on a few private firms, and in my view the transparency is uneven. First Street publishes its methodology. California's regulator now reviews wildfire models before insurers can use them, and Florida has long certified hurricane models through a state commission. But the full workings of the commercial models are proprietary. A homeowner reading a renewal notice usually has no way to see which assumptions moved.

The quiet rule under the bill

For a homeowner, the rule underneath the insurance bill is that the price can move because a model was rewritten, not only because their own risk changed. Sometimes a rewrite reflects real new information, such as another hurricane season or better evidence about how roofs fail. Sometimes it reflects a change in method. From the outside, the two look the same.

The risk map is being redrawn in software, and many people first see the change on their renewal notice.

Sources

Figures current as of October 2026.