What a cat model is actually doing
A catastrophe model runs tens or hundreds of thousands of simulated event years against a property's location, construction, occupancy, and protection (the "exposure data"). For each simulated event, the model calculates wind speed (or fire intensity, ground shaking, flood depth) at the property, applies vulnerability curves to estimate physical damage, and translates damage into dollar loss. The output is not a forecast of next year — it is a probability distribution of loss across many possible years.
Two numbers dominate every cat-model conversation:
- Average Annual Loss (AAL): the expected annual loss if the next 10,000 years played out as modeled. This drives the pure-premium portion of the rate.
- Probable Maximum Loss (PML): the loss at a defined return period (commonly 1-in-100 or 1-in-250 years). This drives capital and reinsurance buying.
The three big modeling vendors and their primary perils
| Vendor | Strong perils | Geographic focus |
|---|---|---|
| Verisk / AIR | Hurricane, severe convective storm, earthquake, wildfire | Global, deep U.S. detail |
| Moody's RMS | Hurricane, earthquake, terror, flood | Global, deep U.S. detail |
| CoreLogic | Wildfire, hurricane, hail, flood | U.S. focused, strong property-level data |
| KCC, ARA HurLoss, ZestyAI, Cotality | Hurricane, wildfire, hail | Specialty challengers used alongside the big three |
How model output drives a homeowner's rate
| Modeled metric | How it shows up on a homeowner's policy |
|---|---|
| AAL increase from a model update | Across-the-board rate filing increase the following renewal cycle |
| 1-in-100 PML increase | Higher reinsurance cost → higher catastrophe load on the rate |
| Distance-to-coast score | Tier-county surcharge, wind deductible structure, eligibility |
| Wildfire score (1 to 100) | New-business eligibility cutoff; declination above carrier's threshold |
| Hail vulnerability score | Roof-cover surcharge, roof-age cap, ACV-only roof endorsements |
| Flood depth at 100-year | NFIP rating now uses Risk Rating 2.0, which is itself a catastrophe model |
What changes when a model is updated
Vendors release new model versions every 1 to 3 years. A major version update is one of the most disruptive events in property insurance pricing because it changes every carrier's view of risk at the same time. Notable industry-wide updates:
- RMS v18 hurricane (2018): raised modeled losses in inland Carolinas, Georgia, and Northeast.
- AIR Severe Thunderstorm v2.0 (2020): raised hail and tornado losses across the Plains and Midwest, contributing to roof-deductible restructuring.
- Wildfire models post-2017 California fires: dramatic increases in modeled wildfire AAL drove the contraction of admitted wildfire capacity that pushed properties to FAIR Plans.
- NFIP Risk Rating 2.0 (2021–2023): moved flood pricing from zone-based to property-specific catastrophe modeling.
Property-level inputs that move the model the most
- Latitude / longitude (geocoded to address, not ZIP centroid).
- Year built and code year (which building code applied when built).
- Construction type (frame, masonry, masonry veneer, steel).
- Roof age, roof shape (hip vs. gable), and roof material.
- Number of stories and square footage.
- Opening protection (impact-rated windows, shutters).
- Wildfire defensible space and ember-resistant features.
- Distance to coast, distance to brush, elevation, slope.
How Agents can help
Agents cannot change the underlying catastrophe model, but they can change how a property scores against it. Agents present the documentation that improves modeled risk — current roof inspection, hip-roof certification, impact-window receipts, wildfire defensible-space photos, elevation certificate — to the underwriter. They also shop multiple carriers because each carrier weighs the models differently and uses different cat-load assumptions, so two carriers can quote the same address 30% to 60% apart on premium for the same coverage.