Insurance Desk

Cat Bond Desk Soren Vaeth

The BSX closing Q2 at a record $70.5 billion in ILS listings is the week's most consequential number for this desk. Cross-reference it against the Artemis dashboard: $65.6 billion outstanding risk capital, $18.9 billion in YTD issuance across 92 deals, average deal size $145 million. The market is not just large — it is structurally deepening. The 5.53% insurance risk spread over a 2.5% market-level expected loss implies a multiple-on-EL of roughly 2.2x. That is not generous by historical hard-market standards, but it is not the 1.5x that characterized the late soft-market complacency of the early 2020s either. Capital is returning, and it is returning at prices that at least nominally compensate for modeled loss — with the collateral yield component (3.76% on T-bills and money-market at effective fed funds of 3.63%) providing a real floor under total return.

The deal flow this week tells a coherent story about where cedents are buying protection. Hannover Re's 3264 Re ($200M, U.S./Canada named storm and earthquake), Swiss Re's Matterhorn Re ($345M, U.S./Canada named storm and earthquake), and the Los Angeles DWP's 123 Lights Re ($100M, California wildfire) are all peak-peril, model-dense transactions. The LADWP deal in particular deserves attention: a public utility transferring California wildfire risk directly into the capital markets, bypassing the traditional reinsurance stack entirely. Post-2023 wildfire losses in California, that is a rational cedent decision — and it signals that the capital markets are willing to take California wildfire at current clearing spreads when the structure is clean.

Tropical Storm Lala introduces a wrinkle. Hawaii hurricane is a real but historically infrequent peril. Most outstanding cat bonds with named-storm coverage in the U.S. will not have Hawaiian wind exposure as a primary driver of their EP curves — the catalog is too thin. If Lala intensifies and makes landfall on the Big Island, the first-order question is not whether a cat bond triggers (attachment probabilities for a single Hawaii event are generally very low relative to Gulf or Atlantic named-storm structures), but whether it forces a reassessment of the historical analog set that underlies Atlantic-Pacific named-storm modeling. That is a longer-cycle concern, not an immediate spread-widening event.

I'd note that my colleague Dr. Chandrasekar's point about thin event catalogs for Hawaii deserves weight here. The spread-over-EL framework gives honest prices only when the EL itself is honest. A peril with five meaningful historical analogs in the record is not the same calibration problem as one with fifty. If Lala makes landfall, it adds to the catalog — but it also proves the model was underweighting a real tail.

At 5.53% insurance risk spread on a 2.5% market EL (roughly 2.2x multiple), ILS pricing is disciplined but not panicked — and the LADWP wildfire deal signals capital markets are absorbing California risk directly at current spreads.

Bias flag — Reads spreads as honest prices; may underweight model error in thin-catalog perils like Hawaii hurricane and systematic underestimation of correlated cyber loss

The Cycle Margaret Ennis

Eighteen-point-nine billion dollars of issuance in 92 deals before mid-August. Last time this desk checked a number like that, rates were hardening and cedents were scrambling for capacity. Now we have Convex — Stephen Catlin's specialty play — reporting 8% premium growth in Q2 2026 with the explicit caveat that prices are sliding. That sentence is the market cycle in miniature: volume is up because cedents are buying more, but price per unit is declining. Capital has come back. The seeds of the next soft market are being planted in the middle of what everyone is still calling a disciplined environment.

Aviva's numbers crystallize the dynamic. A 93.4% undiscounted combined ratio with GI premiums up 29% to £8.093 billion and group operating profit up 24% to £1.326 billion — those are the metrics of a carrier printing money in a market that still has adequate rates. The question the cycle always asks is: how long does discipline hold when results are this good? New capital sees those combined ratios and wants in. The BSX record $70.5 billion in ILS listings is the alt-capital answer to that question: it is already in.

What I am watching is the retrocession layer. Convex growing premiums while prices slide implies that the reinsurers buying retro protection are either paying more for it or taking more net. If retro softens in step with primary specialty, the whole stack compresses together — and when the inevitable large event arrives, there is less buffer at every layer. Lala is a useful early test. Hawaii is not a systemically large market, but the psychology of the first U.S. hurricane landfall of a season matters more than the economics of any one event in determining whether market participants stay disciplined or start convincing themselves that pricing is adequate. Soren's read on the spread-over-EL is correct for modeled risk — where I diverge is on the question of whether the model is capturing the full cost of the next hard market's trigger.

Premium volume is expanding (Convex +8%, Aviva GI +29%) while specialty prices slide — the classic pattern of capital recycling into the late phase of a hardening cycle before the next softening begins.

Bias flag — Mean-reversion lens may be calling 'late cycle' prematurely; structural capital withdrawal (climate non-stationarity, persistent secondary-peril uncertainty) could sustain current pricing longer than cyclical history suggests

Modeled Loss Dr. Ravi Chandrasekar

Tropical Storm Lala presents a textbook secondary-peril / thin-catalog problem. The Artemis report notes this could be the first U.S. hurricane landfall of 2026 — on Hawaii. Let us be precise about what that means for the models. Atlantic and Gulf Coast named-storm models are calibrated against decades of landfall data: tracks, intensities, decay rates, storm surge, inland flooding, construction-class correlations. The central Pacific hurricane basin — which produced Lala — has a categorically different historical record. Meaningful landfalls on the Hawaiian Islands are measured in single digits over the modern observational period. The hazard module for a Hawaii named-storm event in most vendor models is a thin interpolation over sparse data, not a curve derived from observed frequency.

This is the gap that matters: if Lala makes landfall on the Big Island as a Category 1 or 2, modeled loss estimates will carry enormous uncertainty intervals. The actual loss will be determined by factors the model handles poorly — construction vintage (Hawaii has significant older residential stock), post-event demand surge on an island with constrained supply chains, and tourism-sector business interruption that most property models do not explicitly capture. My calibration flag is relevant here in reverse: I would ordinarily caution against over-trusting the EP curve. In this case, the EP curve itself is the uncertainty — the catalog is too thin to have generated a well-converged curve in the first place.

Soren on the Cat Bond Desk is right that attachment probabilities for Hawaii in most outstanding named-storm structures will be low. But that is a function of the model — if the model systematically underweights Hawaii hurricane frequency because of catalog thinness, then the attachment probabilities are not conservative, they are simply artifacts of an underfitted hazard surface. The LADWP 123 Lights Re transaction is a better-modeled deal: California wildfire has a rich recent event catalog and significant vendor model investment. Hawaii hurricane does not have that luxury. The experiment Lala is about to run may add one more data point to a catalog that has needed it for a long time.

Hawaii hurricane is a thin-catalog peril where model EP curves carry structural uncertainty; if Lala makes landfall, modeled loss estimates will have wide confidence intervals driven by demand surge and island supply-chain constraints, not just hazard calibration.

Bias flag — Interrogates EP-curve uncertainty rigorously but underweights social inflation and litigation-driven loss development, which are the primary drivers of reserve deterioration on the casualty side visible in the PRU and Travelers 10-K rewrites

Carrier Books Theo Marchetti

Two carrier prints this week that belong in the same frame. Aviva: group operating profit £1.326 billion, up 24% from £1.068 billion in H1 2025; total GI premiums up 29% to £8.093 billion; undiscounted combined ratio 93.4%. Convex: gross premiums written $1.9 billion in Q2 2026, up 8% year-over-year, across both insurance and reinsurance. These are not the same type of carrier — Aviva is a diversified UK group with life, health, and GI; Convex is a focused specialty platform — but the shared signal is hard to miss: disciplined underwriting is still producing profit, and the scoreboard is green.

The macro backdrop matters here. With VIX at 14.63 (down 4.14 points over 30 days), HY OAS tight at 2.71%, and a normal-positive 10Y-2Y curve of 0.51pp at effective fed funds of 3.63%, the investment income environment is benign-to-supportive. At 3.63% fed funds, carriers with short-duration investment portfolios are earning real yield on collateral — which is exactly what the Artemis block confirms (3.76% collateral yield embedded in the 9.29% cat-bond market return). Combined-ratio improvement and investment income working together in the same direction is the best possible carrier environment. The risk is that everyone knows it, which is the Cycle reading that Margaret is making.

The 10-K novelty data from the SEC filings block adds a layer worth noting. Insurance sector leaders showed 30.3% average Risk Factor novelty (Item 1A) and only 28.3% average MD&A novelty — the second-lowest MD&A novelty of any sector in the corpus. Berkshire Hathaway led with 73.5% MD&A novelty (45.4% Risk Factors), and Travelers led on Risk Factors at 47.2% (246 new sentences added against 251 removed — a near-wholesale rewrite). PRU led on Risk Factor novelty at 66.8% with 304 sentences added. Low sector-average MD&A novelty suggests most insurance leaders are not dramatically reframing their performance narrative — consistent with a stable, profitable environment. But Travelers' Risk Factor rewrite is a flag worth tracking: 88 sentences of net change in how a major P&C writer is describing its risks is not cosmetic.

Aviva's 93.4% combined ratio and £1.326B operating profit (up 24%) alongside Convex's 8% premium growth confirm carrier profitability remains strong, but Travelers' near-wholesale Risk Factor rewrite (47.2% novelty, +246/-251 sentences) is the anomaly that deserves a follow-up call.

Bias flag — Combined-ratio focus underweights long-tail reserve development risk; 93.4% undiscounted is a different number than the discounted equivalent, and the 10-K novelty signal on PRU/Travelers deserves more than a footnote

Solvency Watch Eleanor Pryce

The cyber signal from this week deserves more attention than it is getting in an ILS-heavy news cycle. A hacking group claims mass data theft from nearly 50 companies including Fiserv and GE — and Insurance Journal flags the story from the attacker's own posting. The independent model read correctly labels this Contested: no victim company confirmation is in the corpus. But the solvency and regulatory significance of a breach at a financial infrastructure node like Fiserv — which processes payment transactions for thousands of financial institutions — is not contingent on which direction the attribution resolves. The cyber insurance market has been quietly hardening its exclusions and sublimits for systemic/correlated cyber events since the early 2020s. A confirmed mass exfiltration event of this scale would test those exclusions in ways that have not been litigated at scale.

From a balance-sheet perspective, the more immediate solvency story is the one embedded in Aviva and Convex's results — not because they are stressed, but because good results in the current environment mask the reserve development question. Aviva's 93.4% combined ratio is undiscounted — a meaningful qualifier. The discounted ratio, which incorporates investment income on reserves, will look better; the question is whether reserve adequacy is keeping pace with inflation on open claims from prior years, particularly on U.K. casualty lines where social inflation dynamics are beginning to mirror the U.S. Theo notes the low MD&A novelty for the insurance sector aggregate, but PRU's 66.8% Risk Factor novelty — 304 sentences added — and Travelers' 47.2% rewrite are the stress signals hiding inside that average. A carrier substantially rewriting how it describes its risks in its 10-K is communicating something to regulators and rating agencies before it communicates it to the market.

The unverified Fiserv/GE mass data-theft claim is a stress test for systemic cyber coverage exclusions; PRU's 66.8% and Travelers' 47.2% Risk Factor novelty rewrites are regulatory early-warning signals the sector average obscures.

Bias flag — Treats every Risk Factor novelty spike as a potential distress signal; Travelers and PRU may be updating risk language proactively rather than reactively