Repricing Shows Up in the Spread. Deterioration Shows Up in the Terms.

Cahal Doris, Chief Investment Officer ILS at Twelve Securis, told Artemis this week what most of the market has been saying quietly since the summer: with record cat bond issuance, a broader investor base and a light catastrophe year so far, capital will press on rates at the January 1 renewals. His line on returns was blunt. "As additional capital enters the market, spreads can tighten and the exceptional returns experienced post-2022 should not be regarded as a permanent benchmark."

That part is arithmetic. The part worth reading twice is the one about what to watch as it happens. "Attachment points, covered perils, subject-business definitions and contractual clarity are fundamental components of risk," he said. "The key is therefore to distinguish between normal repricing as market conditions evolve and deterioration in the underlying quality of the risk being assumed." And a warning aimed at his own peers: managers "need to demonstrate that they are not simply accepting market consensus on terms, particularly where pricing becomes more competitive and terms may come under pressure."

We read that as a data problem, and it is one we live with every day.

Two things move in a soft market, and only one of them is easy to see

Repricing is one number. The spread on a new issue is printed in the pricing announcement, tracked by every broker, and comparable across deals to the basis point. When a hurricane bond that would have paid 9 percent over the money-market rate last year clears at 7 this year, nobody has to hunt for that fact. The market sees it on the day.

Deterioration is different. It is a lower attachment on a bond with the same name as last year's. It is a named-storm definition that quietly picks up a named-storm-plus-flood wording. It is a subject-business definition that grows from a personal-lines book to "all property business written by the ceding company and its affiliates." It is an aggregate cover with a franchise deductible that got smaller. None of that shows up in the spread, and most of it lives in the offering circular, which is a few hundred pages that get read carefully once, by the people pricing the deal, and then filed.

Doris' point is that the second kind of movement is where a soft market does its damage, because the spread can look like it repriced 15 percent for the same risk when the risk moved too. Telling the two apart requires holding the terms as data, side by side, across years. Not as PDFs.

What we hold, and what we refuse to hold

Calchis tracks 245 active catastrophe bonds, $65 billion of principal, 60 of them issued in 2026. By trigger type the book is 162 indemnity, 76 industry-loss and 7 parametric. By peril it is 138 multi-peril, 77 hurricane, 24 earthquake and 6 wildfire.

Here is the number that matters for Doris' argument: of those 245, we hold a published attachment point for seven. Every one of the 245 is marked low-confidence for live trigger modeling, and that is deliberate. Our structures registry, the only path by which a bond becomes something the proximity engine will evaluate against a live event, requires each structure to be transcribed from the bond's own offering circular with a page citation, reviewed in a pull request against the cited source, and validated in the test suite before it can be applied. The registry has to be empty rather than guessed at. An earlier version of this product seeded plausible zones and thresholds for a set of bonds without citations, and we purged it, because a wrong attachment point presented with a confident interface is worse than a blank field. A blank field tells the reader to go to the document. A fabricated field tells them not to bother.

So when Doris says "contractual clarity" is a component of risk, we agree in the most literal sense we can: it is a column, it is either sourced or empty, and we do not fill it from consensus.

The secondary-peril question is a multi-peril question

He also expects that "quantifying the risk from secondary perils will likely be under scrutiny." More than half of the book we track, 138 of 245, is multi-peril. That is where the secondary-peril argument actually gets settled at renewal: not in whether a bond names severe convective storm, but in whether the covered-perils clause of a multi-peril aggregate lets a season of hail and tornado losses erode the retention before the hurricane arrives. Two bonds with the same headline peril list and the same attachment can carry very different amounts of that risk depending on how the aggregate is defined and what the per-occurrence deductible is. Again: terms, not spread.

Our contribution there is upstream of the contract. For any US location we return a hazard score for each peril the model covers, including hail, tornado, wildfire and winter storm, with the sources and parameters behind it, and we say which perils are not scored at that location and why. That is the input a manager needs to argue with a covered-perils clause: what the secondary-peril load on this subject business actually looks like, county by county, from public federal data, before the wording is agreed.

Digitization is what makes deterioration visible

The last thing Doris asked for is the one we would put first. "Greater digitisation of data throughout the transaction lifecycle can improve scalability without requiring every transaction to become identical." He was talking about processing a record H1 of issuance and maturities without the market seizing up, and that is true. But a structured record of each deal's terms, attachment and exhaustion, covered perils, subject business, calculation agent, aggregate mechanics, is also the only instrument that can answer his own question. You cannot distinguish repricing from deterioration by reading two offering circulars a year apart. You can by diffing two rows.

That record does not exist yet as a market utility, and we are not going to pretend our seven attachment points are it. What we can say is what the discipline looks like from the data side. Every term is either cited to a page or left empty. Nothing is inferred from the sponsor's previous deal. Nothing is filled from "what these usually look like." It is slower, and it means our catalog has a lot of honest gaps in it. In a market that is about to be flooded with capital and consensus, the gaps are the point.

Source: Cahal Doris, Twelve Securis, as reported by Artemis, 6 September 2026. Quotations are his; the catalog counts are ours as of 6 September 2026, and the reading of them is ours.