Bermuda Reinsurers' 85.3% Combined Ratio Rests on a 9% Premium Decline

Fitch's seven-company Bermuda re/insurer cohort earned an 85.3% combined ratio in the first half of 2026, improved from 90% for full-year 2025, with three full points supplied by favorable prior-year reserve development and only 2.8 points consumed by catastrophes (Fitch Ratings via The Royal Gazette, August 2026). Net premiums written fell roughly 9%. Net-income return on equity slipped to 15.7% from 18.6%.

The group comprises RenaissanceRe, Arch Capital, PartnerRe, SiriusPoint, Axis Capital, Everest Group and Hamilton Insurance Group, and all seven booked an underwriting profit for the half. Four of the seven wrote less premium than a year earlier on reduced prices, smaller exposures and non-renewals, and the comparison is depressed further by the absence of the reinstatement premiums that the January 2025 California wildfires generated. Fitch's framing of what comes next is not celebratory: "Underlying underwriting results are expected to continue to deteriorate into 2027, reducing margins, and weakening overall returns" (Fitch Ratings, August 2026).

Backing the Luck Out of the 85.3%

Fitch disclosed the two swing components alongside the headline, which makes the underlying figure recoverable by arithmetic. Add back the favorable development, remove the catastrophe load, and what remains is an approximate accident-year, ex-catastrophe combined ratio. Run it on both periods and the answer is the same number twice.

Component	H1 2026	Full-year 2025
Reported combined ratio	85.3%	90.0%
Catastrophe load	2.8 pts	6.7 pts
Favorable prior-year development	3.0 pts	2.2 pts
Implied accident-year ex-cat combined ratio	85.5%	85.5%

Implied figure is actuary.info arithmetic on Fitch's disclosed components (reported ratio, plus development, less catastrophe load), not a Fitch-published number. It ignores the period mismatch discussed below.

The 4.7-point headline improvement is entirely attributable to a lighter catastrophe quarter-pair and a slightly heavier release. The underwriting engine beneath it produced 85.5% in both windows. That is the single most useful number in the release for anyone setting a 2027 plan loss ratio, and it is the one the trade coverage did not compute.

The Comparison Period Is Doing Work

Fitch benchmarked a six-month result against a twelve-month one. A first half contains no North Atlantic hurricane season; a full year does. Any Bermuda cohort will look better on a January-to-June window than on a January-to-December window for reasons that have nothing to do with underwriting quality, and that seasonal asymmetry is worth several points of combined ratio in an average year.

The like-for-like read sits in Fitch's wider release. Its 18-company global non-life reinsurer group posted an 86.1% combined ratio for H1 2026, improved from 92.7% in H1 2025, with catastrophes at 3.5 points against 10.9 a year earlier (Fitch Ratings, August 2026). Strip the catastrophe load out of both and the ex-cat result moved the wrong way, from roughly 81.8% to 82.6%. A softening market showing up as underlying deterioration is exactly what the price cuts of the last two renewal rounds are supposed to produce.

The Catastrophe Tailwind, Sized

The 3.9-point reduction in the Bermuda cohort's catastrophe load is not a reserving choice or a retro-buying success. It is the industry loss environment. Global insured natural catastrophe losses reached $42 billion in the first half of 2026, the lowest first half since 2020 and 16% below the ten-year average (Swiss Re Institute, August 2026). Severe convective storm alone accounted for $28 billion of that.

Gallagher Re put the half at $46 billion, against $84 billion in H1 2025 and 28% below the ten-year average of $64 billion (Gallagher Re, August 2026). Aon's estimate of at least $47 billion sits in the same band. Three independent loss aggregators converging near half the prior-year figure is a clean read: the improvement was handed to the sector, not earned by it.

Two structural details limit how far that comfort travels. Severe convective storm is a high-frequency, medium-severity peril that mostly lands in primary and low-layer reinsurance retentions, so a $28 billion SCS half does relatively little damage to a Bermuda excess-of-loss book. And the reinsurance cohort's exposure is concentrated in the second half. A first-half catastrophe load of 2.8 points tells an actuary almost nothing about the full-year figure, which is why the site's quarter-by-quarter tracking of the reinsurance cohort keeps treating mid-year margin as provisional.

An ROE That Fell While the Combined Ratio Improved

Return on equity moving in the opposite direction from the combined ratio is the release's most informative contradiction. The Bermuda group's net-income ROE fell to 15.7% from 18.6%, and it did so while shareholders' equity was flat to down at several constituents: SiriusPoint's equity declined 8%, with marginal declines at Arch and Everest. A shrinking denominator normally flatters ROE. This one did not, which means the numerator fell harder.

Fitch cohort, H1 2026	Combined ratio	Premium change	Net-income ROE
Bermuda seven	85.3%	NPW down about 9%	15.7%
18 global non-life reinsurers	86.1%	NPW down 6%, to $73.73B	18.2%
European big four	76.9%	P&C revenue down 9.4%	21.5%

The Bermuda cohort carries the best-looking headline of the three on a per-point basis and the weakest return. Munich Re, Swiss Re, Hannover Re and SCOR converted a 76.9% combined ratio into a record 21.5% average ROE for the half, a gap the site examined in its coverage of the big four's record return against revenue that is already falling. Part of the spread is leverage and asset mix, and part is that a diversified life and health book carries earnings the property-cat cycle does not touch.

For the Bermuda names the mechanism is simpler. Underwriting margin per dollar of premium held roughly flat at the accident-year level, the premium base shrank 9%, and investment income on a portfolio that has already repriced to current yields no longer supplies incremental year-over-year lift. Multiply a static margin by a smaller base and the absolute profit falls. ROE is the metric that notices premium contraction; the combined ratio is not.

Sizing the Volume Effect

The decomposition is worth doing explicitly, because it separates what management controls from what the cycle handed over. An 85.3% combined ratio leaves 14.7 points of underwriting margin per dollar of net earned premium. Hold that margin constant and shrink the earned base 9%, and absolute underwriting profit falls about 9% before a single loss assumption changes.

The reported ROE fell from 18.6% to 15.7%, a relative decline of roughly 16%. Volume alone accounts for something in the region of nine of those points. The remainder splits between the H1-versus-full-year basis difference and the flattening of investment income, where a fixed-income portfolio already marked to prevailing yields no longer delivers the step-up in net investment income that lifted 2023 through 2025 returns across the Bermuda market.

That decomposition has a planning consequence. If the residual is mostly volume and mostly investment-income normalization, then no amount of underwriting improvement recovers it, because the underwriting side is already producing 85.5% at the accident-year level. The only levers left that move ROE without moving the loss pick are capital return and financial leverage, which is why buyback activity and third-party capital fee income are the line items worth watching in the second-half releases rather than the combined ratio.

Inside the 9% Premium Decline

Whether a 9% contraction reads as discipline or as lost share depends on what shrank. The cohort's own disclosures suggest it is not uniform retreat.

Hamilton grew net premiums written 10%, on casualty reinsurance and specialty insurance, and was the clear exception in the group.
SiriusPoint's reinsurance premium fell 8.9% to $336.9 million in Q2 while its insurance and services premium rose 15.0% to $644.6 million (Bermuda Re, July 2026). Core gross written premium still grew 6%.
The 2025 wildfire reinstatement premiums that inflated the prior-year base do not recur, a comparison distortion rather than a volume decision.

Chief executive Scott Egan described the strategy as having "both the capability and agility to target and grow in attractive areas while pulling back where we don't see adequate returns," while noting that market conditions were "becoming more challenging" (Bermuda Re, July 2026). That is the shape the aggregate hides: capital rotating out of property catastrophe reinsurance and into primary specialty, not leaving the platform.

The pricing backdrop justifies the rotation. Guy Carpenter's global property catastrophe rate-on-line index fell 16% across the 2026 renewal rounds, having already dropped 12% at January 1, the steepest annual decline since the late 1990s (Guy Carpenter, July 2026). Against that, dedicated reinsurance capital is projected to reach a record $705 billion in 2026, up from $663 billion at year-end 2025, split between $575 billion traditional and $130 billion third-party (AM Best and Guy Carpenter, August 2026). Supply is growing into falling demand-side pricing, which is the arithmetic that produces both the rate cuts and the voluntary shrinkage.

Where the Three Points of Development Came From

Cohort-level favorable development of 3.0 points, improved from 2.2 points, is a benign aggregate concealing wide dispersion. RenaissanceRe alone contributed 8.2 points of favorable development, driven by lower-than-expected losses on prior property catastrophe events, while Hamilton booked 1.4 points of adverse development in the opposite direction.

That split matters for durability. Property catastrophe releases are the shortest and most self-limiting form of favorable development available to a reinsurer: once a hurricane or wildfire event closes out below the initial estimate, the redundancy is realized and gone. It cannot be repeated on the same accident year, and a cat-light 2026 generates fewer new events to release from in 2027 and 2028. RenaissanceRe's own defensive posture, examined in the site's coverage of its decision to buy more retrocession behind a 72.8% combined ratio, reads consistently with a management team that does not expect the current loss environment to persist.

Casualty is the other side of the ledger and moves more slowly. Everest's core-business combined ratio deteriorated to 90% in the second quarter from 87% a year earlier, and the US casualty reserve question that has driven charges across the sector since 2024 has not resolved. A cohort release built on property redundancy while casualty accident years 2016 to 2019 remain contested is not the same as broad-based reserve strength, a distinction the site has traced through Schedule P development patterns across the second quarter.

War Losses Enter the Reserve Conversation

Fitch flagged roughly $3 billion of industry insured loss from the Iran conflict as a first-half item for the cohort, small against a $42 billion natural catastrophe half but structurally different from everything around it. Broader market estimates put political violence and war losses from the region in a $3 billion to $4 billion range.

War and political violence claims do not reserve like weather. There is no stochastic event catalog, no hazard footprint and no vulnerability curve, because the loss-generating process is a geopolitical decision rather than a physical one. Three properties make the initial pick unusually soft:

Notification lags run long. Assets in an active conflict zone cannot be surveyed, so the transition from incurred-but-not-reported to case reserve depends on when access is restored, not on a claims-handling cycle.
Development is driven by coverage disputes, not by severity drift. Whether a loss attaches to marine war, aviation war, political violence, terrorism or an all-risks property tower with a war exclusion determines which reserve it lands in, and those determinations are litigated over years.
Aggregation crosses classes that a catastrophe model treats as independent. One conflict can trip marine hull war, aviation hull, energy property and political violence covers in the same quarter without a shared physical trigger.

The site examined the practical version of this at Pelagos, where $60 million of Middle East conflict losses and a $34 million Qatari gas-plant explosion landed in a single quarter. A $3 billion industry figure spread across a seven-company cohort with an aggregate combined ratio of 85.3% is not a solvency question. It is a reserve-adequacy question that will be answered in 2028 and 2029, on accident years being closed today at estimates nobody can validate yet.

What the Cohort Implies for the 2027 Pick

The pricing signal and the reported signal are pointing in opposite directions, and the reported one lags. Rate reductions of 12% at January and a cumulative 16% across the year are earning into the 2026 and 2027 underwriting years, not the results just published. Fitch's expectation of continued underlying deterioration into 2027 is a statement about that lag, not a forecast of catastrophe activity.

An actuary building a 2027 plan from this release gets three usable anchors. The accident-year ex-catastrophe run rate is near 85.5%, and it has not improved in eighteen months despite two years of rate that was supposed to still be adequate. The catastrophe load of 2.8 points is roughly 40% of what the same cohort absorbed in 2025 and should be treated as a draw from the favorable tail of the distribution rather than as a new expected value. And three points of favorable development, concentrated in short-tail property at a single carrier, is not a repeatable input to a forward loss pick.

Load a normalized catastrophe budget back onto 85.5%, hold development at zero, and add the earned-through effect of a 16% rate reduction on the property-cat share of the book, and the arithmetic runs toward the mid-nineties before 2027 begins. The seven Bermuda names appear to be reading it the same way, which is the most credible reading of a 9% premium decline in a year when capital is at a record and nobody is capital-constrained. Shrinking on purpose while margin still looks strong is the behavior of underwriters who know what their own rate change is, and the combined ratio will be the last place it shows up.

Sources

Reduced catastrophe losses lift reinsurance first-half results, The Royal Gazette, August 21, 2026 (Fitch Ratings Bermuda cohort data).
P/C reinsurance net premiums written fall 6% in H1: Fitch, Business Insurance, August 2026.
First-half 2026 insured catastrophe losses: below trend, rising risks, Swiss Re Institute, August 2026.
Gallagher Re estimates global insured cat losses at $46bn for H1'26, Artemis, August 2026.
Guy Carpenter Global Property Catastrophe Rate-On-Line Index, Artemis, July 2026.
Dedicated reinsurance capital expected to hit record $705bn in 2026, Reinsurance News, August 2026 (AM Best and Guy Carpenter).
SiriusPoint CEO Scott Egan warns of tougher market as reinsurance book shrinks, Bermuda Re, July 31, 2026.
Bermuda re/insurers post strong H1'26 results as cat losses ease: Fitch, Reinsurance News, August 2026.
Iran War Could Raise Exposures for Global Terrorism, Political Violence Underwriters, Insurance Journal, March 2026.