Catastrophe modeling
Risk · Cloudbreak team · Updated October 29, 2026 · 4 sources
Definition
A catastrophe model is a computerized process that simulates thousands of plausible catastrophic events, estimates the damage each would do to a portfolio of properties, and applies the policy terms to calculate the insured loss.1
How it works
Catastrophes are rare, so a few decades of claims say little about what the next big hurricane, earthquake or hailstorm could cost. A catastrophe model fills that gap by simulating a very large set of events that could plausibly happen, running each one over the properties insured, and adding up the losses. Most models have four parts, though their names vary:
- Hazard: a catalog of simulated events. Each has a strength, size, location, path and annual probability, plus a footprint of local intensity such as wind speed.
- Vulnerability: damage functions that turn local intensity into expected damage, based on the building's construction, occupancy and height.
- Exposure: the portfolio itself, with its locations, risk characteristics, insured values and policy terms such as deductibles, limits and reinsurance.
- Financial: applies deductibles, limits, attachment points and reinsurance to each simulated event, then combines the results into a probability distribution of losses.
Damage is often stated as a mean damage ratio: the expected loss as a percent of value at a given wind speed or ground motion. A standard deviation shows the uncertainty around it.
The American Academy of Actuaries says actuarial methods built on 5 to 25 years of catastrophe losses proved inadequate for these perils.2
The two main outputs are average annual loss (AAL), the loss averaged across all simulated years, and the exceedance probability (EP) curve, which ranks simulated losses by how likely each is to be exceeded. Probable maximum losses for set return periods are read off the EP curve.
Example
The American Academy of Actuaries ran a hypothetical book of 100,000 Florida homes through a hurricane model. Each home had a $207,500 building value and a 2% deductible. The Academy priced hurricane wind cover as Premium = AAL + expense load + risk load, using a 27% expense load and a risk load based on the standard deviation of modeled losses.2
| County | Modeled loss (AAL) | Premium | Premium on a $207,500 home |
|---|---|---|---|
| Monroe | $13.82 | $56.81 | $11,788 |
| Miami-Dade | $7.60 | $31.25 | $6,485 |
| Hillsborough | $0.75 | $3.09 | $642 |
| Orange | $0.36 | $1.48 | $306 |
| Statewide | $2.64 | $10.86 | $2,254 |
The same model priced mitigation. Hurricane shutters cut a Monroe County home's modeled hurricane wind AAL from $2,872 to $2,479, a 13.7% drop, which lowered the premium 4.5%, or $534. The Academy notes that these figures are illustrations, not realistic loss estimates.2
Where you'll see it
- Exposure data requests: models need each location's address or coordinates, total insured value and building characteristics. A location matched at street, building or parcel level carries less uncertainty than one matched at postal code or city level. Some missing fields are filled with default assumptions.
Limits
Results can vary widely between models, even with the same exposure data. A version update can also change one model's results dramatically.
Hurricane and earthquake models are more mature than newer ones such as wildfire. Severe convective storm models work from an incomplete record, because storms in sparsely populated areas often went unrecorded. They must also reflect how hail, tornadoes and straight-line winds each add to losses.3
Some core assumptions are proprietary. Damage that comes with an event but falls outside the model's output has to be handled separately.2
State rules
Questions
- What are catastrophe models used for?
- Insurers and reinsurers use them for ratemaking, setting premium credits for mitigation, buying reinsurance, and assessing capital and solvency.
- Do catastrophe models predict the next hurricane?
- No. The models estimate the probability of losses from events that could occur. Their event catalogs cover what is possible, but they do not predict future events.
- Which perils do catastrophe models cover?
- Natural catastrophes such as hurricanes, earthquakes, floods, convective storms (including tornadoes and hail) and wildfires. Models also cover man-made catastrophes such as terrorism, and emerging risks such as cyber.
- Why do two catastrophe models give different answers?
- Models differ in their data specifications and underlying assumptions. Companies often adjust the output or combine several models to get a range of outcomes.
Cloudbreak
Cloudbreak keeps every policy, statement of values and loss run in one place per property, tracks renewals, and watches every building against storms with its proprietary predictive storm engine, built with AI and backed by meteorologists.
Sources
Written by the Cloudbreak team. Numbers, dates and rules cite the published sources below.
- 1. National Association of Insurance Commissioners (NAIC): Catastrophe Models (Property)
- 2. American Academy of Actuaries, Extreme Events and Property Lines Committee: Uses of Catastrophe Model Output (July 2018)
- 3. National Association of Insurance Commissioners (NAIC): NAIC Catastrophe Modeling Primer (adopted August 13, 2025), Attachment Three to Executive (EX) Committee and Plenary minutes
- 4. Florida Commission on Hurricane Loss Projection Methodology: Hurricane Standards Report of Activities as of November 1, 2023