Cat Bond Spreads Reprice Within Days of a Major Disaster — Damage Estimates Take Weeks to Months to Catch Up

Summary

Catastrophe-bond spreads move almost immediately — within a single trading day to the next weekly secondary-market pricing cycle (roughly 1-7 days) — while the actual, confirmed damage from a disaster typically takes weeks to months to firm up. The bond market's first move is not a reaction to confirmed losses at all: it is a forward-looking repricing driven by the storm's forecast track and catastrophe-model (RMS/AIR) nowcasts, and in the most extreme case (Hurricane Irma, 2017) the largest single-day move in the Swiss Re Cat Bond Price Return Index's history (-16%) happened a full two days before landfall, purely on the threat. Magnitude scales with modeled exposure and uncertainty: whole-market index drops have ranged from about -1.3% (Milton, 2024, a glancing/lower-than-feared hit) to -16% (Irma threat) to a roughly 5-10% broad mark-to-market decline (Ian, 2022), with individual exposed tranches falling far further (many Florida-exposed Irma bonds were marked down to 20-75 cents on the dollar in a single day). By contrast, FEMA's own federal record of the same events shows initial disaster declarations issued in 1-9 days, but the fuller confirmation of damage — additional counties/areas added to a declaration as damage assessments catch up — takes 30-60+ days, and industry-wide insured-loss estimates (RMS/AIR) and cat bond price convergence back to a fundamentals-based value take weeks (Milton: ~3 weeks) to, in the worst cases, years (Irma indemnity-triggered bonds remained below par two years later). So the bond market leads the damage-estimate process by roughly an order of magnitude in speed, then gets revised again as real numbers arrive.

How fast the spread move happens

Cat bond secondary-market pricing sheets get marked down essentially in real time as a storm's track and intensity forecast develop — before any wind or water has touched ground. On Hurricane Irma (Sept 2017), the Swiss Re Global Cat Bond Price Return Index fell 16% in a single day (Sept 8, 2017) — the steepest one-day decline in the index's history — purely because Irma's forecast track threatened South Florida's densely insured value; the storm hadn't made landfall yet. Individual Florida-exposed tranches (Blue Halo Re, Citrus Re, Oak Leaf Re, Casablanca Re) were marked down that same day to between 10 and 75 cents on the dollar by brokers.

On Hurricane Milton (Oct 2024), the market's own weekly pricing cadence shows the lag was a single pricing cycle: the cat bond market yield stood at 11.01% the Friday before landfall (Oct 4), Milton made landfall Oct 9, and by the very next pricing (Oct 11) the yield had jumped to 11.70% — a roughly 6% relative jump reflecting spread widening on repriced positions. The Swiss Re index's broad measure fell 1.34% at that same first post-landfall pricing (US wind-specific sub-index -3.64%), well below the ~15% loss some earlier scenario forecasts had priced in.

On Hurricane Ian (Sept 2022), pricing sheets were marked down within days of landfall to imply a market-wide 5-10% decline (roughly $1.6bn-$3.6bn of value), with some individual tranches suffering near-100% principal loss as it became clear Ian was a market-changing event.

How fast damage estimates actually firm up

Querying askamerica's disasters.disaster_declarations table (FEMA OpenFEMA) directly for Hurricane Ian and Hurricane Milton shows the gap: FEMA issues an emergency declaration (EM) within 1-2 days of an incident's official start — essentially a formality that authorizes federal resources, not a damage figure — and the first Major Disaster declaration (DR) at 6 days. But that first DR only covers an initial set of counties. The fuller picture, as more counties'/areas' preliminary damage assessments (PDAs) are completed and added to the declaration, arrives much later: for Ian, an amended DR-4673 adding designated areas was issued Nov 21, 2022 — 59 days after the storm began; for Milton, a second DR (4844) was issued Nov 5, 2024 — 31 days after. This matches FEMA's own published PDA process, under which a state generally has up to 30 days to complete initial damage assessment collection, and non-expedited major-disaster-declaration requests take an average of about 13.4 days to decide once submitted — i.e. the federal damage-confirmation process is inherently a multi-week-to-two-month process, not a same-week one.

On the reinsurance-industry side, RMS and AIR (Verisk) issue rapid initial modeled loss estimates within days of landfall, but these are themselves revised for weeks as claims data comes in, and periodically rebased on new model versions (e.g. the RMS V23 hurricane model update in spring 2024 moved cat bond spreads independent of any live storm, with the top-10 most-repriced cat bonds all showing expected-loss increases of at least 21%, several over 50%—demonstrating that spread moves track evolving model-based loss views, not just realized claims). Actual insured-loss convergence takes far longer: after Irma, industry-index-triggered cat bonds recovered to par within 217 days, but indemnity-triggered bonds (whose payout depends on the sponsor's own realized claims, not a modeled industry index) remained below par even two years later.

Putting the two clocks side by side

The bond market therefore reacts roughly an order of magnitude faster than the federal/insurance damage-confirmation process — it prices the modeled, forecast-based risk almost instantly, then gets revised (usually partially reversing the initial move, as in Milton and the broader post-Irma/Ian pattern) as real loss data and RMS/AIR model updates arrive over the following weeks to months.

Every query behind this report

1 warehouse call ran in this session, in order. Each is reproducible against the same snapshot.

SELECT DISTINCT disaster_number, declaration_type, declaration_title, incident_begin_date, declaration_date FROM disasters.disaster_declarations WHERE declaration_title ILIKE '%MILTON%' OR declaration_title ILIKE '%IAN%' ORDER BY declaration_date