For three consecutive years, hail, tornadoes, straight-line winds and derechos produced more than USD 45 billion in insured losses annually, eclipsing hurricanes as the dominant driver of U.S. catastrophe losses and accumulating more than USD 200 billion in three years. The reclassification of severe convective storms from "secondary" to "primary" peril is not a semantic upgrade. It is the recognition that a structural shift in the U.S. loss landscape has already occurred, driven by exposure growth, construction cost inflation, and a hazard environment that cat models, calibrated to historical data and designed for single-event analysis, were not built to capture.
The three-year period from 2023 to 2025 represents not an unusual cluster of bad luck but the crystallisation of a structural shift in the U.S. catastrophe loss landscape. Between 2023 and 2025, losses accumulated to a total in excess of USD 200 billion, according to Gallagher Re. The U.S. is the number one SCS hotspot, accounting for more than 80% of the value of insured losses globally. Cumulative losses from severe convective storms now exceed those from hurricanes, challenging the traditional classification of these events as secondary perils.
The reclassification of severe convective storms, encompassing tornadoes, hailstorms, straight-line winds, and derechos, from secondary to primary peril is the most significant categorical shift in the U.S. insurance loss landscape since the recognition of earthquake risk following the 1994 Northridge earthquake. It reflects not merely a change in the physical hazard but a convergence of four mutually reinforcing trends: expanding urban exposure in storm-prone areas, sharply rising construction and repair costs, social inflation increasing claim settlement values, and a growing body of evidence suggesting that climate change is expanding the geographic range and seasonal window of SCS activity.
This case study differs structurally from others in this series, which focus on single landmark events. SCS in 2023–2025 is examined as a multi-year accumulation of losses, an aggregation of hundreds of individual events, none of which alone would qualify as a peak-zone catastrophe, but whose cumulative financial impact rivals or exceeds the largest single-event catastrophes in U.S. history. This distinction, between single catastrophic events and sustained high-frequency attritional loss, is itself one of the most important analytical features of the SCS problem for cat modellers and reinsurers.
The term "secondary peril" originally referred to hazards that were secondary in both frequency-weighted impact and peak-loss magnitude. SCS was classified as secondary because individual events rarely caused multi-billion-dollar losses and were considered attritional rather than catastrophic. The 2023–2025 data has demolished both assumptions. SCS events now routinely generate multi-billion-dollar individual losses, and their cumulative annual totals exceed those of primary perils. The terminology has not caught up with the reality, and where it persists, it actively misleads underwriters and capital allocators about the true risk.
Ice particles formed within convective updrafts, ranging from pea-sized to grapefruit-sized. Roof damage is the single largest driver of SCS insured losses. A single severe hailstorm can affect a metropolitan area simultaneously, generating dense claim concentrations impossible to anticipate spatially.
Violently rotating columns of air rated EF0–EF5. Despite dominating media coverage, tornadoes generate a smaller share of total SCS insured losses than hail, but produce catastrophic local concentrations. EF4–EF5 events can completely destroy entire communities within a narrow path.
Non-rotating downburst winds, including derechos, long-lived organised straight-line wind events that can affect hundreds of miles simultaneously. The May 2024 Houston derecho brought 100 mph winds to the nation's fourth largest city, causing more structural damage to high-rises than Hurricane Beryl did six weeks later.
Convective precipitation can produce flash flooding in urban areas where impervious surfaces prevent infiltration. Lightning-ignited fires and direct strikes to structures and electrical systems contribute additional losses, secondary drivers within SCS events but primary in specific geographic contexts.
The United States, particularly the corridor from Texas to the Canadian border known as Tornado Alley, and the expanding Dixie Alley across the southeastern states, experiences the most frequent and intense SCS activity in the world. This is a direct product of North America's unique geography: the Rocky Mountains create dry air that descends over the western Great Plains forming the dry line; the Gulf of Mexico provides unrestricted warm, moist northward airflow; strong upper-level jet stream winds provide the vertical wind shear necessary for supercell formation; and flat terrain allows storms to develop and travel hundreds of miles without disruption.
The spring tornado season, typically March through June, corresponds to the period when Gulf moisture, Rocky Mountain dry air, and jet stream wind shear align most frequently. Activity is rising: Cotality reported 142 days of hail of two inches or greater across the country in 2025, against a 20-year average of 122.
Research over the 2010s and 2020s has documented a systematic eastward shift in tornado activity, with Dixie Alley, Mississippi, Alabama, Tennessee, and the Carolinas, experiencing increasing tornado frequency and a higher proportion of nocturnal and EF3+ events. Dixie Alley has higher population density than the traditional Tornado Alley plains, older and less tornado-resilient housing stock, a higher proportion of mobile homes, and a greater proportion of nocturnal events, when residents are asleep and warning system effectiveness is lower.
Indicative attribution based on industry research synthesis; contributions vary by event and region
Urbanisation and development into hazard-prone areas have dramatically expanded exposure. Oklahoma City's building stock increased by an estimated 40% and its developed land area by 60% between 1990 and 2025. This reflects a national pattern in which the fastest-growing metropolitan areas are disproportionately located in the SCS-prone South and Midwest, Dallas-Fort Worth, Houston, Nashville, Kansas City, and Indianapolis have all experienced rapid suburban expansion into areas that have always been exposed to hail and tornado risk but previously contained far less insured value. Hailstorms now pose a threat to 41 million homes at moderate or greater risk, representing a reconstruction cost value of USD 13.4 trillion. For tornadoes, 66 million homes are at risk, valued at USD 21 trillion RCV.
Roof replacement is the single most common and most expensive SCS insurance claim. The cost of asphalt roof replacements has reportedly surged 250% since 2000, with a 45% increase in the last five years alone. Supply chain disruptions, skilled labour shortages, and aging infrastructure further amplify costs. This cost inflation is not fully captured in standard insurance policy limits, which are typically set at inception and updated infrequently, producing systematic underinsurance that mirrors the Turkey earthquake building code gap: the policy says one thing, the economic reality at time of loss says another.
Social inflation, the increase in insurance claims costs attributable to litigation trends, jury award escalation, plaintiff attorney strategies, and shifting legal interpretations, has materially increased SCS loss costs. In the SCS context it manifests primarily through public adjuster proliferation (third-party firms that maximise hail damage claim payments), roofing contractor referral networks driving claim filing rates well above the underlying physical damage rate, Assignment of Benefits litigation in some states, and nuclear verdicts in commercial SCS liability cases.
The climate change contribution to the SCS loss trend is the most complex and contested of the four amplifiers, the physical science is not settled in the way that the link between warming sea surface temperatures and tropical cyclone intensification is settled. Climate modelling suggests that total hail frequency may decrease while large hail (≥5 cm) frequency increases; there is evidence of a poleward and eastward shift in tornado activity expanding the Dixie Alley pattern; and increasing atmospheric moisture provides more fuel for convective storms. Climate risk models from CoreLogic and Cotality suggest that by 2030 and especially by 2050, regions like the South and Midwest will face even greater exposure to hail, tornadoes, and high-wind events.
| Date | Year | Event / Location | Loss (USD) | Defining Feature |
|---|---|---|---|---|
| Mar 2–3, 2023 | 2023 | Southern Plains / SE Outbreak | USD 6.1B | 33 confirmed tornadoes; 70–90 mph winds hit Dallas-Fort Worth metro; opening shot of record 2023 season |
| Mar 24–27, 2023 | 2023 | Rolling Fork EF4, Mississippi | USD 1.9B | 17 fatalities; 195 mph winds; entire Rolling Fork business district destroyed; majority-Black low-income community; Dixie Alley pattern |
| Mar 31–Apr 1, 2023 | 2023 | Central / Eastern Tornado Outbreak | USD 4.3B | 145 tornadoes across Midwest and East; third billion-dollar SCS event in 30 days, unprecedented sequence density |
| Jun 14–19, 2023 | 2023 | Southern / Plains Outbreak + Perryton EF3 | USD 3.5B | 92 tornadoes; EF3 kills 3 in Perryton TX; 4.75" hail near Caledonia MS; 1M+ outages |
| May 6–10, 2024 | 2024 | Super Outbreak - 180 Tornadoes | USD 6.1B | 180 confirmed tornadoes across 13 states; EF4 in Barnsdall OK at 180 mph; 6.25" hail in Johnson City TX; largest single outbreak of 2024 |
| May 16, 2024 | 2024 | Houston Derecho | USD 1.2B | 100 mph straight-line winds hit Houston, more structural damage to downtown high-rises than Hurricane Beryl caused two months later; paradigm-shifting urban derecho event |
| May 25–27, 2024 | 2024 | Memorial Day Outbreak | USD 3.5B | 97 tornadoes overnight; 15+ fatalities; EF3 in Eddyville KY; Greenfield IA EF4, 5 fatalities; high nocturnal casualty pattern |
| Sep 24, 2024 | 2024 | Oklahoma City Hailstorm | est. >USD 2B | 35,000 homes damaged in a single day, most impactful single hail event of 2024; illustrated concentrated urban hail exposure |
| March 2025 | 2025 | 26-State March Outbreak | USD 8–10B | Largest single SCS event of the three-year period; first EF5 tornado in 12 years; early-season extreme violence across 26 states demonstrated expanded seasonal risk window |
| May 18–21, 2025 | 2025 | Great Plains / Mid-South Outbreak | USD 2.6B | 133 tornadoes; EF3 in Iuka KS at 160 mph; 4.5" hail near Arnett TX; second major outbreak of 2025 within 8 weeks of March event |
"Secondary cat perils that are now becoming primary cat perils, such as wildfire and tornado hail, have contributed to record natural catastrophe losses. At a 500-year return period, modelled losses from a single extreme hailstorm reach USD 58 billion, hurricane-scale exposure from a peril most portfolios still treat as secondary."
— Moody's / Allianz Commercial / J.S. Held analysis, 2025–2026Of all individual events in the 2023–2025 period, the Rolling Fork EF4 of March 24, 2023 warrants particular attention for what it reveals about vulnerability patterns in the expanding SCS loss geography. Rolling Fork is a community of approximately 1,800 people in Sharkey County, Mississippi, a rural Delta community with high poverty rates, majority-Black population, and a building stock dominated by manufactured homes, older wood-frame structures, and tenant housing.
On the evening of March 24, a large, violent, long-tracked multi-vortex wedge tornado struck Rolling Fork, Silver City, and Midnight. The tornado killed 17 people and injured at least 165 others, with peak winds of 195 mph over a path length of 59.4 miles and width of 0.75 miles. The entire business district on Highway 61 was destroyed.
The Rolling Fork event illustrates several SCS-specific risk dimensions distinct from hurricane and earthquake risk: mobile homes are approximately 15–20 times more likely to be destroyed in a given tornado than site-built homes at equivalent wind speeds, and are disproportionately concentrated in the SCS-prone rural South; when a tornado destroys the business district and a significant proportion of the residential stock of a community of 1,800 people simultaneously, reconstruction requires external resources and is measured in years; and as with Hurricane Maria and the Kahramanmaraş earthquake, Rolling Fork's pre-existing poverty and inadequate housing stock amplified the impact of a physical event that, in a more affluent community, would have been devastating but survivable.
The May 16, 2024 Houston derecho demonstrated something not clearly established in the insurance record: that a non-tornadic convective wind event, with no rotating supercell structure, could produce concentrated losses in a major metropolitan area comparable to a weak hurricane. A derecho with winds of up to 100 mph occurred in Houston, causing significant damage and at least eight fatalities. The storm caused more structural damage to downtown Houston's modern high-rise buildings than Hurricane Beryl, which made direct landfall on Houston as a Category 1 hurricane six weeks later.
The explanation lies in the specific combination of wind direction persistence, building aerodynamics, and the element of surprise. The derecho's sudden onset meant that building management had not taken preparatory actions that would reduce hurricane exposure. The straight-line, sustained wind direction also produced different loading patterns on building facades than the rotating wind signature of a hurricane. Post-event engineering surveys found glazing failures on specific building faces loaded most severely by the wind direction, a lesson for both building design and SCS vulnerability model development.
SCS presents cat modellers with a set of challenges genuinely different from hurricane, earthquake, or flood modelling. A hailstorm's damage footprint can be as small as a few square kilometres, requiring much higher-resolution modelling than hurricane or earthquake hazard. There is no single peak event, unlike hurricanes, where cat models can identify clear scenario events, SCS risk is characterised by hundreds of events per year with highly variable individual severity. Reliable spatially detailed SCS loss data is available only for the past few decades. And according to Moody's, SCS events are highly localised and frequent, with historical data often inconsistent, making traditional modelling approaches unreliable.
Most cat models are designed around the concept of a single catastrophic event. SCS is fundamentally an accumulation problem: the insurance industry's concern is not any single hailstorm but the total of all hailstorms across an annual season. Traditional catastrophe excess of loss reinsurance is structured around individual occurrence limits, it responds to a single large event but does not aggregate across the multiple smaller events that collectively drive SCS losses. The result is that much of the SCS loss falls below Cat XL attachment points and is retained by primary insurers as attritional loss, the direct driver of the personal lines market hardening across the South and Midwest.
The most fundamental challenge for SCS cat models is the same one identified for wildfire models in the LA 2025 case study: the historical data that calibrates the model is from a world with different exposure values, different construction costs, and potentially different SCS hazard frequency and severity than the world in which losses will actually occur. Losses continue to escalate, driven primarily by increasing property exposure, urban sprawl, and rising replacement costs and inflation. A model calibrated to historical SCS losses will systematically underestimate current and future SCS losses unless it explicitly accounts for exposure growth, cost inflation, and climate trends, which most legacy SCS models did not do adequately entering the 2023–2025 period.
| Dimension | Severe Convective Storms (SCS) | Atlantic Hurricanes |
|---|---|---|
| Annual insured losses (US, 2023–25 avg) | >USD 50 billion annually | Variable, USD 0–100B+ depending on landfalls |
| Cumulative 2023–25 US insured losses | >USD 175 billion | ~USD 50–80 billion (no major US landfall 2023–25) |
| Warning time | Minutes to hours (tornado); little for hail | Days to a week for hurricane track and intensity |
| Geographic footprint | Highly localised (km-scale); multiple simultaneous events | Regional (hundreds of km); typically single event |
| Primary loss driver | Hail (50–80% of losses); then wind, tornado | Wind (traditional); increasingly surge in major events |
| Reinsurance response | Mostly retained by primary insurers below Cat XL attachment | Typically triggers Cat XL and catastrophe bonds |
| Cat model maturity | Historically underdeveloped; rapid advancement post-2023 | Mature, well-calibrated; multi-vendor competition |
| Peril classification 2020 → 2025 | Secondary peril → Primary peril | Primary peril → Primary peril |
Cumulative SCS losses now exceed hurricane losses over the same period, challenging the traditional secondary peril classification. The reclassification drives changes in capital allocation, reinsurance treaty design, pricing methodology, and the priority assigned to SCS model development investment. Cat model vendors have responded with substantially enhanced SCS capabilities, including Moody's RMS US SCS HD Model released in late 2025.
Legacy SCS models relied primarily on historical storm track data and statistical relationships between storm characteristics and losses. Post-2023, investment in physics-based convective storm simulation, resolving individual supercell dynamics, hail trajectory, and wind field structure at sub-kilometre resolution, has accelerated substantially. Moody's RMS US SCS HD Models integrate high-resolution, physics-based hazard simulations with the largest-ever claims-calibrated vulnerability dataset, capturing over USD 55 billion in industry claims.
The recognition that SCS risk is fundamentally an accumulation problem has driven development of annual aggregate SCS models that explicitly simulate the full distribution of seasonal loss rather than treating each event independently. These models are essential for reinsurance treaty design, capital modelling, and the pricing of aggregate stop-loss and annual aggregate reinsurance structures.
The industry-wide shift to percentage-based wind and hail deductibles represents a fundamental restructuring of the risk-sharing relationship between insurers and policyholders for SCS losses. By creating a meaningful homeowner retention on the first loss, percentage deductibles reduce claim frequency for moderate events while preserving coverage for catastrophic losses, but transfer significant financial risk to households, particularly lower-income households who may not have reserves to self-insure a USD 8,000–15,000 deductible.
Post-2023, building codes in several high-SCS states have been updated to require impact-resistant roofing in new construction. The IBHS FORTIFIED program, certifying construction meeting enhanced wind and hail resistance standards, has seen dramatically accelerated adoption. Engineering research has demonstrated that Class 4 impact-resistant shingles reduce roof damage from severe hail by 70–90%, the most cost-effective single mitigation measure for residential SCS losses.
The rapid expansion of solar farms and data centres in storm-prone areas has introduced new vulnerabilities. Solar panels are acutely vulnerable to hail, the glass and cell structure is not designed for impact resistance. As the U.S. solar fleet expands rapidly across the SCS-prone South and Plains, the aggregate hail exposure of the energy sector is growing at extraordinary speed. Specialist SCS vulnerability functions for solar installations are being developed and incorporated into commercial cat model platforms.
This event is directly referenced in Exposure Data Preparation for Catastrophe Modelling - the Intermediate course on The Return Period - in Lesson 4.4 (Emerging Risks and Non-Modelled Items), where the 2023–2025 SCS period is used as the primary illustration of a secondary peril reclassifying as primary; the SCS accumulation problem as a fundamental mismatch between single-event cat model frameworks and frequency-driven attritional risk; and exposure non-stationarity as a cause of systematic legacy model underestimation.