Skip to content
Tuesday, October 27, 2026 · Today’s edition, 6:30 a.m. ETTue, Oct 27 · 6:30 a.m. editionSign inMy Desk 0EditionsSign in

New cat bond spreads fall to 2.4 times modeled loss

Expected loss rose even as spreads fell, shifting due diligence to the models behind each bond.

At a glance

  1. In the first half of 2026, expected loss rose even as spreads fell, the authors write, leaving the 2026 multiple roughly 35% below its 2024 level.
  2. For an allocator holding cat bonds, directly or through a fund, the thinner multiple shifts due diligence from the coupon to the model.
  3. Many models used in the US are built from backward-looking data, the authors write.

New catastrophe bonds are being priced at 2.4 times modeled expected loss so far in 2026, down from 3.71 times in 2024, according to a Woodwell Climate Research Center analysis published Oct. 20 by CAIA.

In the first half of 2026, expected loss rose even as spreads fell, the authors write, leaving the 2026 multiple roughly 35% below its 2024 level.

For an allocator holding cat bonds, directly or through a fund, the thinner multiple shifts due diligence from the coupon to the model. The authors describe the multiple as a cushion, and a smaller one leaves less room for model error. The question is whose model produced the expected loss figure and whether its baseline reflects the climate the bond will actually face.

Model vintage is the blind spot

Many models used in the US are built from backward-looking data, the authors write. Hurricane models in particular are built from storms recorded between 1900 and today, producing a statistical replica of that record rather than a view of what comes next.

The investors buying the risk and the insurers originating it do not use the same climate assumptions. The authors argue that a spread quoted as a multiple of modeled expected loss describes the model as much as the peril, and two bonds carrying similar language can be underwritten to different views of the same storm.

That leaves allocators with questions about the denominator: which model, which version of its event set, and whether expected loss is measured against a backward- or forward-looking baseline. A thinner multiple on one model can still represent a wider cushion than a richer multiple on another. The authors' conclusion is conditional: a thin cushion works only if the models behind it are accurate, making the model rather than the headline spread the thing to underwrite.

Reported by the Insurance-Linked & Specialty desk from CAIA Association. Published . Updated Oct 24, 2026 · 4:09 p.m. ET. Corrections

Related

Alternatives

AMP and One River call for a defensive system, not a single hedge

The post questions whether government bonds still reliably hedge inflation.

Alternatives

CAIA says AI could flatten single-family office hierarchy

It proposes an information steward accountable for the data agents read.

Alternatives

Bitcoin's crypto market share hits 60% as ETF flows waver

CoinDesk reported the share at its highest since May; smaller-token indexes fell and ETF flows stayed volatile.

Alternatives

AM Best affirms Kenya Re at B as 2025 combined ratio hits 110.6%

The stable outlook rests on balance-sheet strength AM Best calls very strong, while the Kenyan reinsurer's risk-management capabilities are rated weak.

Alternatives

777 Partners cofounder Josh Wander pleads guilty to $500m fraud

Sentencing is March 1; Bermuda's regulator had ordered 777 Re to shed affiliated assets.

Alternatives Weeklyfrom Private Wealth Daily

Manager research and fund selection for wealth allocators.

Free. Every morning. Unsubscribe from any edition in one click.

From other sections

New cat bond spreads fall to 2.4 times modeled loss