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 development and only 2.8 points consumed by catastrophes (Fitch Ratings via The Royal Gazette, August 2026). Net premiums written fell roughly 9%, and net-income return on equity slipped to 15.7% from 18.6%.
Key Takeaways
- Back the luck out and the underwriting engine produced 85.5% in both windows. The 4.7-point headline improvement is entirely a lighter catastrophe load and a slightly heavier release.
- Global insured natural catastrophe losses reached $42 billion in the first half, the lowest since 2020 and 16% below the ten-year average. The improvement was handed to the sector, not earned by it.
- ROE fell while the combined ratio improved, because a static margin multiplied by a base 9% smaller produces less absolute profit. ROE notices premium contraction; the combined ratio does not.
- Fitch's own forward view is not celebratory: underlying results are expected to continue deteriorating into 2027, reducing margins and weakening returns.
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.
| 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% |
The implied figure is arithmetic on Fitch's disclosed components rather than a Fitch-published number, and it ignores a period mismatch worth naming. 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 looks better on a January-to-June window for reasons unconnected to underwriting quality.
The like-for-like read sits in Fitch's wider release. Its 18-company global non-life group posted 86.1% for the half against 92.7% a year earlier, with catastrophes at 3.5 points against 10.9. Strip the catastrophe load from 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 two renewal rounds of price cuts are supposed to produce.
The catastrophe tailwind is an industry loss environment rather than a reserving choice or a retro-buying success. Global insured losses reached $42 billion in the half, with severe convective storm alone at $28 billion. Gallagher Re put it at $46 billion against $84 billion a year earlier, 28% below a ten-year average of $64 billion, and Aon's estimate of at least $47 billion sits in the same band.
Two structural details limit how far that comfort travels. Severe convective storm is high-frequency and medium-severity, landing mostly in primary and low-layer retentions, so a $28 billion SCS half does relatively little damage to a Bermuda excess-of-loss book. And the cohort's exposure is concentrated in the second half, so a first-half load of 2.8 points says almost nothing about the full year.
An ROE That Fell While the Combined Ratio Improved
Return on equity moving opposite to the combined ratio is the release's most informative contradiction. Net-income ROE fell to 15.7% from 18.6%, and it did so while shareholders' equity was flat to down at several constituents, with SiriusPoint's equity declining 8% and marginal declines at Arch and Everest. A shrinking denominator normally flatters ROE. This one did not, which means the numerator fell harder.
For the Bermuda names the mechanism is simple. 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 already repriced to current yields no longer supplies incremental year-over-year lift. Multiply a static margin by a smaller base and absolute profit falls.
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 constant and shrink the earned base 9%, and absolute underwriting profit falls about 9% before a single loss assumption changes.
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 period-basis difference and the flattening of investment income, where a fixed-income portfolio already marked to prevailing yields no longer delivers the step-up that lifted 2023 through 2025 returns.
That has a planning consequence. If the residual is mostly volume and mostly investment-income normalization, no amount of underwriting improvement recovers it, because the underwriting side is already producing 85.5% at the accident-year level. The levers left that move ROE without moving the loss pick are capital return and financial leverage.
That is why buyback activity and third-party capital fee income are the second-half line items worth watching rather than the combined ratio. The contrast with the European big four, who converted a 76.9% combined ratio into a record 21.5% ROE, is partly leverage and asset mix and partly a diversified life and health book the property-cat cycle does not touch.
Inside the 9% Premium Decline
Whether a 9% contraction reads as discipline or as lost share depends on what shrank, and the cohort's own disclosures suggest it is not uniform retreat. Hamilton grew net premiums written 10% on casualty reinsurance and specialty insurance. SiriusPoint's reinsurance premium fell 8.9% to $336.9 million in the second quarter while its insurance and services premium rose 15.0% to $644.6 million, with core gross written premium still growing 6%.
Part of the decline is not a volume decision at all. The January 2025 California wildfires generated reinstatement premiums that inflated the prior-year base, and those do not recur, so a portion of the 9% is a comparison distortion rather than business walked away from.
What remains is the harder read. Four of the seven wrote less premium on reduced prices, smaller exposures and non-renewals, which is the behavior a disciplined market is supposed to produce. But discipline and lost share look identical in a premium line, and the distinguishing evidence, whether the business declined was priced below the cohort's own loss pick, is not in any public disclosure.
That leaves the 85.5% accident-year figure carrying more weight than it comfortably can. It is stable across both windows, which is reassuring on the surface, and it sits on a book that shrank 9% into a softening market while Fitch expects underlying results to deteriorate into 2027. A margin that holds flat while the premium supporting it contracts is not the same signal as a margin that holds flat on a stable base, and only the second one survives another two renewal rounds unchanged.
Further Reading
- Fitch: Big Four Reinsurers' Record 21.5% ROE Masks Price Cuts Still to Earn Through
- RenaissanceRe Buys More Retro Behind a 72.8% Combined Ratio
- Everest Q2 2026: An 88.5% Treaty Combined Ratio in a Softening Reinsurance Market
- Pelagos's 99.5% Combined Ratio Hides a $122 Million Reserve Swing
- What Schedule P Says About Q2 2026 Reserve Releases
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.