The smart money stopped looking at how many hurricanes are forecast a long time ago.
On May 21, the National Oceanic and Atmospheric Administration's (NOAA) official outlook called the 2026 Atlantic season below normal: 55 percent probability, 8 to 14 named storms. But by the close of trading three days later, the reinsurance, cat bond, and prediction markets were all pricing a different season — one that costs more to insure, more to hedge, and harder to ignore.
What the risk markets have figured out is that the much-publicized hurricane season storm count is decoupled from the actual damage a single storm can inflict.
A Category 5 hitting Miami in a "quiet" year — 1992's Hurricane Andrew is the canonical case, or 2024's Hurricane Helene, which stalled inland and caused an estimated $78.7 billion in damage — produces more insured loss than a busy season full of named storms that miss land.
Four signals, four different seasons
NOAA.
The May 21 outlook from NOAA's Climate Prediction Center is the only forecast in the set that draws a single number-line: 55 percent below-normal probability, 35 percent near-normal, 10 percent above-normal. The central call is for 8 to 14 named storms, 3 to 6 hurricanes, and 1 to 3 majors. The long-run average is 14 / 7 / 3. NOAA Lead forecaster Matt Rosencrans anchored the call on an expected El Niño development tightening wind shear over the Atlantic basin, which is a key suppressor of hurricane formation.
The RMN Hurricane Dashboard launches alongside today.
The dashboard tracks the analogs of the four hurricane/market risk signals at the daily level: NHC active storms and outlook, hurricane-exposed equity, the cat bond and listed P&C capital tape, and prediction-market activity.
As the season moves, the question is not which signal turns out to be right — it is where the gap between them widens, compresses, or breaks.
Reinsurance
Capital is proving more abundant than catastrophe risk. According to Aon, global reinsurance capital reached a record $785 billion at the end of 2025, up nearly 10 percent year-over-year, driven by strong retained earnings, investment gains, and record levels of alternative capital.
That capital overhang is translating directly into softer pricing. According to Howden, risk-adjusted global property-catastrophe rates fell 14.7 percent at the January 2026 renewals, while property retrocession pricing declined 16.5 percent, the sharpest reduction in property-cat pricing since 2014. That's despite another year of elevated catastrophe losses, including the Los Angeles wildfires, competition for catastrophe business remains intense, and reinsurers continue to deploy capital aggressively.
Capital in the global reinsurance market now stands at approximately $785 billion, supported by both traditional reinsurer balance sheets and a record $136 billion of third-party capital. Some retreat from secondary perils — wildfire and severe convective storm in particular — remains evident, but not from peak U.S. hurricane risk, where abundant capacity, lower retrocession costs, and strong reinsurer profitability have intensified competition for well-structured business even as attachment points remain higher than pre-2023 levels.
Cat bonds.
The Swiss Re Global Cat Bond Total Return Index has delivered three consecutive years of double-digit returns, gaining 11.4 percent in 2025 and continuing to post positive performance through the opening months of 2026. According to Swiss Re, investor demand remained sufficiently strong that secondary-market spreads tightened throughout much of 2025 despite record issuance and growing outstanding market volume.
At the same time, catastrophe bond issuance has continued to outpace maturities, keeping net capital flowing into the sector and pushing outstanding cat bond risk capital to new highs.
Prediction markets.
Kalshi's 2026 Atlantic named-storm markets currently imply roughly 13 named storms and six hurricanes — at the upper end of NOAA's forecast range of 8 to 14 named storms and 3 to 6 hurricanes.
Meanwhile, Polymarket's contract on whether a Category 4 hurricane will make U.S. landfall before 2027 is trading around 35 percent, implying roughly a one-in-three probability of a high-end landfall event this season.
The prediction markets are assigning relatively little probability to either an exceptionally quiet or exceptionally active season, instead clustering around the upper end of NOAA's range. Volume is concentrated in the over/under contracts nearest NOAA's forecast boundaries — a clean signal of where retail and semi-institutional money believes the binary action will be.
The market is effectively pricing a near-normal season overall while still attaching meaningful odds to a tail-risk landfall event.
The question is why they disagree
The natural first explanation is different information.
Maybe the cat bond professonals know something NOAA's models miss, or the reinsurance underwriters have proprietary modeling that adjusts the forecast. Two senior voices at one of the three largest catastrophe modeling firms make clear it isn't that.
Wesley Terwey, senior scientist at Verisk Catastrophe and Risk Solutions, frames the upper bound on what any seasonal forecast can deliver:
Seasonal forecasting explains about 60 to 70 percent of the season... the remaining 30 to 40 percent of the variance is explained by local factors, things that occur on more local and regional scales.
That is not Verisk's number — Terwey is citing Dr. Phil Klotzbach at Colorado State, whose own outlook is one of the inputs. The point: the formal forecasting apparatus knows where its claims end.
Peter Sousounis, formerly vice president and director of climate change research at Verisk (AIR Worldwide), goes further. Asked how Verisk's internal seasonal outlook is constructed, he was direct:
My recollection is that there was no real Verisk view per se. Rather, it was a comment that they expect the season to be somewhere within the range forecasted by the ensemble of agencies. In that sense it was rather anti-climactic given that interested persons could essentially formulate that result given available public information.
If even the cat modelers, whose business is interpreting catastrophe risk for the firms that price it, do not have a proprietary seasonal view, the markup that reinsurance and cat bonds charge over actuarial expected loss cannot be an information-asymmetry premium. It has to be something else.
It is.
Adam Solomon, Assistant Professor studying public finance and financial economics (with applications to climate and insurance) at NYU Stern and whose recent paper "When Insurance Markets Fail: Catastrophe-Risk Frictions and Public Reinsurance" decomposes the wedge between market price and expected loss, identifies the structural answer. Reinsurance has historically charged approximately 70 percent over expected losses; cat bonds up to 200 percent. Capital cost — the return demanded for holding any tail risk — explains some of that gap. But the bulk falls into two specific buckets: correlation and ambiguity.
The correlation problem is what makes Atlantic hurricane risk uniquely expensive to insure. Solomon says
To the extent that we define the pool of capital as catastrophe risk... then there is just a limit to how much diversification a Florida hurricane can find. It is going to always be very big... All the NATCAT risk is held by the same group of people — and there's a limit.
The reinsurers, ILS funds, and listed P&C insurers that hold US-wind exposure all hold it together. There is no second market of buyers — pension funds, sovereign wealth funds, BlackRock-scale capital — that can absorb correlated Florida wind without already being exposed to it. ILS was supposed to bridge that gap. Solomon's diagnosis is that the bridge isn't holding up:
The hope was always that with the ILS and cat bonds, the pool of risk would not just be natural catastrophes — it would be the global equity and global bond markets... But it just doesn't seem to be the case.
What is changing is the buyer base. Cat bonds, once "kind of exotic," now sit in university endowments, multi-strategy hedge funds, and alternative-manager allocations that did not touch the asset class five years ago. That mainstream-ization adds capacity. It does not eliminate the underlying correlation — every new buyer is still exposed to the same Florida wind every existing buyer is.
The ambiguity premium is what is left. Even after accounting for capital cost and correlation, markets charge for the parts of catastrophe risk that cannot be modeled at all. Terwey's 30 to 40 percent of unmodeled variance lives here and Solomon's describes how that uncertainty compounds:
If you get the one parameter wrong — like, 'oh, our model says there's going to be three hurricanes a decade,' and it turns out to be four — you're 25 percent off for the entire book. So your mistakes just get magnified when you're dealing with these systemic risks, one way or the other.
This is why a "below normal" outlook does not translate into below-normal pricing. The forecast bounds what the modelers think the season looks like. The markup is for what they know they cannot see.
Which brings us back to 1992.
Andrew formed in a quiet season — six named storms, well below the long-run average — and produced what was, at the time, the single most expensive insurance event in US history. Terwey, asked how seasonal forecasts relate to loss outcomes, reached for a comparison:
You can win the grand prize with a small number of raffle tickets, but it's a little less chance. If you have more raffle tickets, you're more likely to win at least something — and you're also more likely to potentially get the big one.
Count and cost are not the same metric. The number of storms in a season tells you the probability of any given storm. The intensity of any one of them tells you how much that storm costs when it lands.
Markets price the second. Forecasters publish the first. Both are doing exactly what they are designed to do.
What to watch
Adam Solomon's paper has emerged as a leading academic case for federal public reinsurance, and the U.S. Senate is reviewing legislation along the lines he describes. Solomon's explains:
I think it'll end up looking like the banking sector — where the government doesn't do much direct banking, but they do deposit insurance and they do lending of last resort. The FDIC and the Fed are insuring the tail risk there... not trying to be in the game of actual day-to-day risk management.
How and where the private market moves in an out of US hurricane risk remains the bigger question. And what role and risks new forms of capital, like prediction markets, bring to the table had yet to be explored. Solomon adds:
There was a report of somebody bet on temperature in Paris and was then found at the weather station using a hair dryer in an apparent attempt to manipulate the sensor. I thought that was quite interesting... If it's wind speed, I don't know how the wind speed gets measured, but maybe you just need a very powerful fan to manipulate that sensor up.
