Severe Thunderstorm Went From $4 Billion to $40 Billion a Year. The County Is the Unit Now.

Harry White, Head of Commercial Strategy at Property Claim Services, gave Artemis the numbers in July: average annual severe thunderstorm insured losses in the United States went from approximately $4 billion a year in 1998 to 2005 to over $40 billion a year in 2021 to 2025. Claims went from 1.3 million a year to 3.5 million. Texas alone was 95 percent of first-quarter 2025 catastrophe claims, its share of total insured losses up from 13 to 23 percent, its annual count of designated events from six to 32. His conclusion was aimed at ILS: "understanding the changing profile of this peril is crucial for those structuring and pricing this risk."

The other half of the story is where Verisk is taking the index. The Verisk Catastrophe Index takes PCS's state-level industry loss and, using the modeling side of the house across "millions of individual locations," breaks it down to the county. The stated purpose is that county-level bonds can be "more accurately modelled and therefore priced." Hurricane first, then tropical storm, earthquake, winter, wildfire, severe weather, tornado and hail.

Put those two facts together and the unit of the market changes. A peril that used to be priced as a state aggregate is now the largest line, and the index that settles it is moving to the county. The county is also, not by coincidence, the unit of American emergency management.

The same polygon, two readers

A county emergency manager's jurisdiction is a polygon. An index bond's coverage region, once the index is county-resolved, is a set of those polygons. During a severe weather outbreak both readers want the same thing at the same time: which warnings are inside the boundary right now, what is in there, and what changed since the last look. The emergency manager wants it to decide about shelters and low-water crossings. The fund wants it because the first data point an index-linked position will ever get is the warning, and the industry loss figure that eventually settles the bond comes later, and is revised after that.

We built the county side first, for the emergency manager, and the second reader turned up on their own.

What we hold at county resolution

Every U.S. county has a hazard history page built from the federal record, counted by county, with each figure linked to the row it came from. Miami-Dade County, Florida reads: 1,394 NOAA storm events recorded from 1996 to 2026, 43 FEMA disaster declarations naming the county since 1965, 61,977 NFIP flood claims paying $852.6 million. The page also says what it cannot tell you — that NOAA files some events against a forecast zone and not a county, that damage is in nominal dollars, that a claim count is not a rate.

Live, an NWS warning is drawn on the footprint the Weather Service actually issued it for: the union of the counties or zones the product names, not a pin at a state centroid. Since mid-September the convective products are split into hail and tornado rather than filed together as "severe weather," which matters when the question is which of the two eroded an aggregate. Every monitored area — a county, a city, a drawn polygon — is evaluated against the live feed once a minute for the conditions its owner cares about, and a condition that is met stays on the screen until someone acknowledges it.

That is the same county polygon, read live, with the same public data an index will eventually be settled on, except earlier and without the dollars.

What we do not hold

We hold no industry loss number. We do not derive one from warnings, hail cells, claim counts or anything else, and the 76 industry-loss bonds in our catalog are marked as not evaluated for exactly that reason. A warning is a warning; PCS's figure is PCS's figure. What we can say is which counties were under which products, when, from the federal record, at the moment a duty officer or a fund manager needed to know it. As the market moves its settlement to the county, that is the part of the picture that is public, and it is the part nobody was watching in real time.

Sources: Harry White, PCS (Verisk), as reported by Artemis, 20 July 2026; Artemis, "Further details about the Verisk Catastrophe Index". The Miami-Dade figures are from our county page, counted from the federal record on 15 September 2026.