Marsh published its third-quarter 2026 Global Insurance Market Index on October 15, and the composite number looks like relief for buyers: global commercial rates fell 5%, a ninth consecutive quarterly decline, though at a slower pace than the second quarter’s 6% drop. It was the first time in two years the pace of softening slowed, according to Commercial Risk’s coverage of the release. Property rates fell 10% globally and cyber fell 4%.
The US detail is where the reserve story sits. US excess and umbrella casualty rates rose 13% in the quarter, per the index figures reported with Marsh’s third-quarter earnings, down from 15% in the second quarter and 18% in the first. US workers’ compensation rates fell 5%. Global casualty rose 1%, and the US is essentially the entire reason the line is rising at all. That spread of 18 percentage points between what excess carriers charge and what workers’ comp carriers charge is the widest the index has logged this year, and this was the first quarter where it stopped narrowing.
Trade press reads this as a pricing story. Read against the loss-cost data, it is a reserve story, and the workers’ comp side of the divergence is the trap.
Two markets, two severity trends
Start with what the excess carriers are reacting to. Chubb CEO Evan Greenberg said on the company’s second-quarter earnings call that US casualty loss costs are rising 6 to 7% annually for primary and 9.5 to 12% for excess, adding there is “zero evidence across the industry that loss costs have abated.” Lockton’s October 2026 market update shows the same split in medians: lead umbrella pricing climbed 6.2% in the second quarter and excess casualty 8.2%, while general liability rose just 1.8%. In liability, pricing is running below where loss costs are heading.
Now the workers’ comp side. Carriers cut rates 5% in the same quarter. They can do that because calendar-year results look excellent: the National Council on Compensation Insurance (NCCI) reported in its updated State of the Line release on October 14 that private insurers posted a final 2025 calendar-year combined ratio of 90.6%, the 12th consecutive year of underwriting profitability, on net written premium of $41.6 billion.
But the accident-year picture is different. NCCI’s 2025 accident-year combined ratio was 102%, and the gap between the two numbers is the whole warning: favorable development on old accident years is subsidizing current pricing. Meanwhile indemnity and medical lost-time claim severity each rose an estimated 4% in accident year 2025, lost-time frequency fell 2%, and in the first half of 2026 the direct loss ratio rose two percentage points to 52% even as direct written premium fell 0.4%.
So the two markets are behaving rationally toward two different realities. Workers’ comp carriers are cutting price off calendar-year profits while the claims they write develop at 4% severity. Excess carriers are raising price into a 9.5 to 12% loss cost trend. The mistake a self-insured can make is treating either renewal quote as a loss-cost estimate.
Who it affects
Self-insured employers with large workers’ comp deductibles, retro-rated programs, or a captive funding both workers’ comp and liability retained layers; public entity pools and JPAs carrying both lines; hospital captives with professional liability towers above a self-insured retention (SIR); and any buyer whose renewal letter shows a workers’ comp credit while the balance sheet carries the claims anyway. If your program retains the first dollar of both workers’ comp and general liability, this divergence runs straight through your funding.
Where this shows up in your reserves
The lever is trend selection, the severity and frequency assumptions that drive your expected claim ratios. Open your last workers’ comp actuarial report and find the trend page. If the selected severity trend sits below NCCI’s 4% figure, ask why, in writing. Frequency and severity need to be checked separately: the 2% frequency decline supports lower expected claim counts, but 4% severity means every open claim costs more at each valuation, and IBNR (incurred but not reported) on open claims compounds both.
For the excess and umbrella layers above your retention, find the trend used to develop excess-layer IBNR. If it is below the 9.5 to 12% range the excess market is pricing, your layer funding and attachment adequacy assumptions lag the claims the market expects. We covered the calendar-versus-accident-year gap and what it does to recent-year margins in NCCI’s 102% accident-year signal, and the mechanics of trend-driven IBNR in what’s driving your IBNR higher.
Also look at the latest diagonal of your incurred triangles. With the accident-year combined ratio above 100, recent accident years carry less margin than the calendar-year result implies, so adverse development on 2024 through 2026 accidents will surface there first. For captives, the same check applies to the funding-versus-indicated comparison your board reviews; see tail factor selection for captives for how excess-layer tail assumptions interact with funding, and our earlier piece on excess casualty pricing versus long-tail development.
What this means for your next review
Put trend selection on the agenda as its own item, deliberately separated from the renewal quote letter. Ask your actuary to benchmark the workers’ comp severity trend against NCCI’s 4% and the excess-layer trend against the 9.5 to 12% range, and to document the basis for any selection below those marks. Then ask the structural question: if your attachment point has not moved, how much of the 13% excess increase is pure loss-cost trend versus a tightening in the excess carrier’s own risk appetite? A flat attachment with rising excess price implies carriers expect more claims to pierce the attachment, which is an attachment-adequacy finding, not just a premium line item.
Checklist for the next 30 to 90 days
- Pull the trend assumptions page from your most recent actuarial report and split frequency from severity; benchmark severity against NCCI’s 4%, not your renewal quote.
- Ask your actuary whether the excess and umbrella IBNR trend selection reflects the 9.5 to 12% loss-cost range, and to document why if it does not.
- Review the latest diagonal of your incurred triangles for case strengthening or late reporting on 2024 through 2026 accident years, where margin is thinnest.
- Confirm with your broker how much of the 13% excess increase is trend versus appetite, and whether your attachment point should move before the January 1 renewal.
- Calendar the Q4 Marsh index, due mid-January 2027, and NCCI’s next State of the Line update against your reserve study and renewal dates.
The read into 2027
The deceleration from 18% to 15% to 13% suggests the excess increases are cresting, not reversing. With loss costs still running 9.5 to 12% in the excess layer and Marsh expecting soft reinsurance conditions to persist through the January 1 renewals, high-single-digit to double-digit excess increases into 2027 are the base case. Workers’ comp softness likely persists a while longer, but NCCI’s own 2026 combined ratio projection of 91% to 96% and the first-half loss ratio uptick point the same direction as the accident-year 102% figure: the old-year subsidy burns off eventually. The reserve risk concentrates in the retained workers’ comp layer, where quotes fall and severity rises.
An independent reserve review brings a second pair of eyes that’s free of the TPA’s or fronting carrier’s incentive structure. We’re working on a directory of independent reviewing actuaries. If you’d like to be considered, get in touch.
Sources
- Marsh Global Insurance Market Index, Q3 2026
- Commercial Risk: Softening continues in Q3 but pace slows: Marsh (October 16, 2026)
- Business Insurance: Workers comp medical, indemnity severity each up 4%: NCCI (October 14, 2026)
- Insurance Business: Casualty loss costs outpace pricing as commercial market splinters (October 2, 2026)
- The Insurer via SquawkNews: Marsh Q3 earnings call coverage, US excess casualty rates up 13% (October 15, 2026)