H2 Hurricane Season Meets the Softest Cat Market Since the 1990s: A Primary Carrier Actuarial Stress Test

The Guy Carpenter Global Property Catastrophe Rate-On-Line Index closed the July 1 renewal cycle down 16% year-to-date and 23% from the 2024 hard market peak, the steepest single-year decline since the late 1990s (Guy Carpenter, July 2026). Swiss Re Sigma 1/2026 projects full-year 2026 insured catastrophe losses at $148 billion; with H1 global losses running below the five-year average of approximately $38 billion, the second half must account for roughly $110 billion to reach trend during peak hurricane season. A market repricing of reinsurance cost does not change that number.

-23%

Global Property Cat ROL from 2024 Peak (Guy Carpenter)

$148B

Swiss Re 2026 Full-Year Expected Insured Cat Losses

$785B

Record Global Reinsurer Capital (Aon, April 2026)

32%

Global ROL Cushion Above 2017 Soft Market Floor

The July 1 Renewal: A Trajectory With No Brake

Dean Klisura, Guy Carpenter's President and CEO, described the mid-year outcome: "Cedents have secured competitive pricing and terms on their reinsurance programs, but many are also exploring alternative options" (Guy Carpenter, July 2026). The phrasing contains two signals. The first is cedent satisfaction: buyers got what they came for on price and terms. The second is structural: primary buyers are redeploying their reinsurance savings into parametric coverage structures, aggregate excess of loss, and supplemental US coverages that the same renewal report documents as growing in demand. When buyers diversify away from traditional treaty layers even as those layers get cheaper, the pricing advantage is being used to restructure cession strategy, not simply lower cost. That behavioral shift has implications for how carriers model their retained risk positions.

The July 1 data closed a trajectory that accelerated at each 2026 renewal date without interruption. The Guy Carpenter US Property Cat Rate-On-Line Index fell 12% at January 1 (the deepest single-renewal decline since 2014), extended to 16% US year-to-date by mid-year, and the APAC index reached 19% below its year-start level by July. The global index sits 23% below the 2024 hard market peak and 32% above the 2017 soft market floor. The reinsurance capital base that enabled every step of this decline is $785 billion (Aon, April 2026): traditional capital at $649 billion, up more than 8% over 2024 year-end, and alternative third-party capital at roughly $136 billion, up 18%. The outstanding cat bond market exceeded $63.9 billion at end-March 2026, a record, with $15.8 billion in new issuance across 60 deals through H1 (Guy Carpenter, July 2026). Cat bond spreads fell more than 20% year-on-year. Every segment of the risk transfer chain repriced lower in the same direction.

The H2 Stress Test: Swiss Re's $148 Billion and the Variance Already Used

Swiss Re's Sigma 1/2026 frames the second-half actuarial problem precisely. The long-run expected annual insured catastrophe loss trend for 2026 is $148 billion, and the report is explicit that below-trend results in any single period represent favorable variability, not a structural revision to the underlying hazard. The 2025 total came in at $107 billion, 28% below that trend line and notable for producing zero Atlantic hurricane landfalls for the first time in a decade. Swiss Re described that outcome as favorable variability. The same framing applies to the first half of 2026: Q1 global insured losses totaled approximately $20 billion, 26% below the 10-year quarterly average (Gallagher Re, April 2026), and H1 losses have tracked below the five-year first-half average of approximately $38 billion.

Metric	Amount	Source
Swiss Re 2026 full-year expected insured losses	$148 billion	Swiss Re Sigma 1/2026
2025 actual global insured losses	$107 billion	Swiss Re Sigma 1/2026
Q1 2026 global insured losses	~$20 billion	Gallagher Re, April 2026
H1 five-year average (for comparison)	~$38 billion	Multiple broker reports
Implied H2 2026 losses to reach trend	~$110 billion	Derived from above
Swiss Re peak-loss scenario for 2026	~$320 billion	Swiss Re Sigma 1/2026

The arithmetic produces a second-half expected loss requirement that is large relative to the H1 actuals. This is not a forecast that Q3 will be catastrophic. It is the expected value calculation that sits inside every cat model: if $38 billion has been consumed in H1 and the best estimate for the full year is $148 billion, the second half carries $110 billion of expected loss exposure entering July 1. A primary carrier actuary who treats benign H1 results as evidence of a below-trend year and reduces H2 cat load picks accordingly is using favorable variability as if it were a change in expected loss. That is not an actuarially defensible basis.

NOAA's May 2026 forecast projected a below-normal Atlantic hurricane season on the strength of the La Nina transition, and below-normal seasons do carry a lower probability-weighted expected landfall rate. But cat models calibrated to multi-year average activity distributions do not mechanically adopt single-season forecasts as a revision to the annual EAL. Incorporating a below-normal forecast as a reduction to the primary cat load requires an explicit methodology step: a quantified adjustment to the annual activity rate, a documented basis for the adjustment magnitude, and conditions under which it would be reversed. The historical record also constrains how much weight any single season forecast should carry. Hurricane Andrew formed and intensified during the 1992 season, which was classified as below-normal, and produced what was at that time the costliest insured catastrophe in US history.

The Rate Arithmetic for Primary Carriers

The mechanics of connecting a reinsurance ROL decline to a net retained loss cost pick are specific enough to trace through a concrete example. A primary carrier purchasing a $50 million catastrophe layer attaching at $25 million net retained loss, with a 10% starting ROL, pays $5 million in annual ceded premium. A 16% ROL decline reduces that to $4.2 million. The ceded premium cost drops by $800,000, and if the carrier allocates reinsurance cost to its rate indication as a proportion of ceded premium paid, the indicated ceded cost in the rate filing also falls proportionally.

The net retained position does not follow. Above the carrier's retention and below the reinsurance attachment, the carrier holds all ground-up losses. The expected annual loss for that retained band is sized by the cat model's hazard, exposure, and vulnerability modules, none of which have changed because Guy Carpenter's index moved. The model's EAL estimate for retained risk reflects the same wind hazard parameters, the same coastal property exposure, and the same vulnerability curves that it did in January. The reinsurance market has repriced the cost of transferring risk above the attachment; it has not repriced the physical risk that sits below it or within the retained band. An actuary who reduces the net retained cat loss cost pick in proportion to the ROL decline is using the wrong causal chain. Reinsurance pricing moves in response to supply and demand in the risk transfer market. The physical risk does not.

The correct methodology treats the ceded cost allocation and the net retained loss cost as separate calculations that happen to interact at the attachment point. The ceded ROL change flows through the ceded cost in the rate indication. The net retained EAL comes from the cat model, adjusted only for documented changes in the model output, the underlying exposure, or actuarial assumptions about long-run activity rates. Market pricing is a relevant input to the ceded cost allocation; it is not a direct input to the net retained loss cost pick.

The 32% Cushion and When the Adequacy Argument Ends

The 32% premium the Guy Carpenter Global Property Cat ROL Index still carries above the 2017 soft market trough is the context that separates the current market from an unambiguous adequacy problem. The 2017 floor was the low-water mark of a multi-year soft cycle that followed the loss-light period from 2012 through 2016. When the 2017 hurricane season delivered Harvey, Irma, and Maria with combined insured losses of roughly $100 billion (Swiss Re), that floor was tested and proved inadequate: a seven-year hard market followed.

The current market is softening from a materially higher starting point. That cushion is real. Reinsurers entering the 2026 cycle with rates 32% above the prior trough have more pricing margin before their returns compress to cost-of-capital than reinsurers at the 2017 floor had. At the current pace, roughly 16 percentage points of global ROL decline per year, the cushion narrows to approximately 16% above the 2017 floor by January 1, 2027, without a major loss event. If a significant H2 event occurs this year, that cushion erodes faster and the January 2027 renewal debate shifts from "how far do rates fall" to "do rates stabilize." David Flandro, Howden Re's head of industry analysis, put the structural tension directly: "Capital has rarely been more abundant in an environment of elevated risk exposure" (Howden Re, January 2026).

For primary carrier actuaries, the 32% cushion has a different implication than it does for reinsurers. It describes reinsurer capital adequacy relative to the 2017 trough, not primary carrier rate adequacy relative to underlying risk. A primary carrier's property cat rates may be running below adequacy independently of where the reinsurance ROL index sits. The two interact through the ceded cost allocation but they are not the same calculation. A primary actuary who reasons that because the reinsurance market has pricing headroom above the 2017 floor, primary net rates are therefore adequate, has conflated market-level capital adequacy with carrier-level rate adequacy. The reinsurer's cushion is not the primary carrier's cushion.

Retrocession and the Whole-Chain Repricing

The retrocession data from the January 1, 2026 renewal (the most recent renewal with detailed cross-segment figures) illuminates how far down the chain the repricing has reached. Howden Re reported property cat reinsurance down 14.7% risk-adjusted, retrocession down 16.5%, and direct and facultative pricing leading all segments at 17.5% lower (Howden Re, January 2026). That sequencing means reinsurers' own loss buffers got cheaper at the same renewal where primary carriers' reinsurance costs fell.

In a moderate cat scenario, that is supportive: cheaper retrocession sustains reinsurer balance sheet stability across a frequency of mid-sized events that do not individually breach attachment points, reducing the probability of mid-cycle capital withdrawal that could otherwise tighten primary carrier reinsurance access during peak season. But in a severe scenario, the whole-chain repricing amplifies the shock. If primary rates, reinsurance rates, and retrocession have all moved lower in concert, a Category 4 landfall in Tampa Bay, the Southeast Florida corridor, or the Gulf Coast refinery complex does not encounter a progressively more expensive risk transfer chain that buffers each layer; it encounters a simultaneously repriced chain where margin has been reduced at every level. The outstanding ILS market adds over $63.9 billion in cat bond limit, but cat bond spreads have compressed more than 20% year-on-year alongside the traditional market. Cat bond investors have accepted lower expected returns at the same time traditional reinsurance cheapened. If a major H2 event breaches trigger thresholds, secondary market spread widening in the ILS segment could make additional mid-season capacity expensive exactly when primary carriers need it most.

Net Loss Cost Picks and the Documentation Requirement

The governance question for primary carrier pricing teams as of July 1 is specific. Market conditions have moved reinsurance rates lower. An actuary who incorporates that signal into a net retained loss cost pick that departs from the cat model's expected value output carries a documentation obligation that is more urgent now than it was at January 1, for a concrete reason: the Swiss Re $148 billion annual projection is on record, the Q1 actuals are on record, and the July 1 renewal data is now on record. A net loss cost pick taken between July 1 and September 1 that departs materially from the cat model in a favorable direction will be evaluated against all three of those data points if a significant H2 event makes the assumption look optimistic in hindsight.

The market-conditions argument that supports a favorable departure takes several forms: that competitors are pricing below model indications and the carrier needs to match for retention reasons; that cheaper reinsurance transfers more risk economically and reduces the all-in expected cost on a net basis; that a below-normal NOAA forecast justifies a favorable seasonal adjustment; or that benign H1 results provide evidence of a below-trend year. None of these is inherently impermissible as an actuarial professional judgment. Each requires: a statement of the magnitude of the departure from the cat model output, the reasoning that supports it, the quantitative basis for the adjustment, and the conditions under which the actuary would revise the assumption. Absent that documentation, a favorable market-conditions adjustment in a July 2026 net loss cost pick is an undisclosed departure, not a disclosed professional judgment.

The documentation creates the paper trail that distinguishes an informed, disclosed professional judgment from a backward-looking liability. Actuaries who conduct the H2 stress test now, explicitly modeling combined ratio outcomes under a $50 billion, $100 billion, and $150 billion H2 cat loss scenario, and who document the basis for their net retained loss cost picks against those scenarios, have a defensible record regardless of what the Atlantic produces between now and November 30. The stress test is not the product of pessimism about hurricane season; it is what the Swiss Re expected loss data and the July 1 renewal repricing together require of a P&C actuary operating under professional standards.

The 32% cushion above the 2017 soft market floor is real, and it means the reinsurance market has not yet crossed the line where pricing adequacy is structurally broken. But that cushion is narrowing at 16 percentage points per renewal year, the expected H2 loss requirement is large relative to H1 actuals, and the cat model has not changed because the risk transfer market repriced. Primary carrier P&C actuaries who carry an undisclosed model-to-market departure into the highest-exposure quarter of the year are carrying a position the July 1 data makes harder to defend with each week of hurricane season that passes without a major event.