Hurricane Polo Hit 892mb and Paid Mexico's Cat Bond Nothing

The National Hurricane Center's Advisory 9, issued at 3 PM Central on September 22, put Hurricane Polo stationary 215 miles south of Zihuatanejo with 180 mph winds and a central pressure of 892mb (NHC, September 2026). Mexico's $175 million IBRD CAR Mexico 2024 Pacific cat bond starts paying at 937mb. It paid nothing, because the pressure condition and the location condition were never satisfied in the same hour.

Six days later Polo crossed the coast of Baja California Sur just south of Las Barrancas at 965mb and 110 mph, 28mb above the threshold (NHC, Advisory 34A, September 2026). Artemis confirmed the notes untouched the next morning. Investors were paid a 12% risk margin against a 4.09% expected loss to carry that week, and the forecast ensembles run while the storm stalled show what the margin was carrying.

Key Takeaways

45mb separated Polo's 892mb peak from the 937mb pressure condition, and the storm sat at or below that level for parts of four days offshore; the gate line runs along the coast, and it crossed at 965mb.
2.93 times expected loss is what the 12% margin pays on a 4.09% modelled loss; a 6.26% attachment and 2.54% exhaustion probability imply a 65% conditional severity and a 41% chance a triggering storm takes the full $175 million.
77 of 1,000 ensemble members triggered the bond in Isacco Loconte's first Google WeatherLab run, mean loss near 45%: about 3.5% of principal in expected loss from one storm, before the count fell to 22 and then zero.
$84 million is Mexico's Pacific risk margin over the four-year term, 85% more margin for the same modelled loss than the 2020 notes, which priced at 6.5% and paid $60 million on Otis.
Three near misses since settlement (Erick inside the gate at 939mb, the M7.3 quake 0.1 short, Polo deep enough outside the gate) against zero payouts, about $54.6 million a year of program margin, and a $575 million parametric layer beneath whose status was open at landfall.

The Gate: Pressure and Position Read at the Same Instant

Mexico's Pacific tranche is what the World Bank calls a cat-in-a-gate: lines drawn along the coastline, each segment carrying its own central-pressure thresholds, with a stepped linear payout function (World Bank, February 2025). Artemis puts the Pacific activation at 937mb or lower as the storm's centre passes a zone line, paying 25% of principal at the first step and rising to 100% depending on where the breach occurs, per occurrence, with AIR Worldwide as calculation agent (Artemis, September 2026).

Polo met the pressure half of that test for most of the week. After the 892mb reading it rose to 905mb and then 927mb on September 23, sat between 923mb and 935mb through September 24 and the morning of the 25th, then re-intensified to 908mb that afternoon with 180 mph winds (Artemis, September 2026). By September 27 it had filled to 952mb, 954mb the next morning, and 965mb at the gate. Every reading below 937mb came with the centre offshore, and every zone-line crossing came above it.

Mexico's 2024 redesign chose this geometry on purpose. Mexico "aimed to optimize the structure to close its coverage gap while minimizing basis risk" and "looked for quicker payouts for hurricane events" (World Bank, February 2025). Speed came from the data source: payouts now key off NHC B-deck data available two weeks after the final advisory, replacing the tropical cyclone reports that pushed the Otis payout into April 2024. The geometry stayed, because a line crossing at a reported pressure is the only thing a calculation agent can settle in two weeks.

Hurricane Erick ran the same test from the other side in June 2025. It reached 939mb while heading into the gate near Oaxaca, 2mb above the threshold, and landed at 950mb (Artemis, June 2025). Erick had the location and lacked the pressure. Polo had the pressure and lacked the location. The gate sorted both into zero, and the two storms together mark out the floor of basis risk that a redraw of the lines leaves in place.

Pricing the Week: 12% Against a 4.09% Expected Loss

The Class D notes settled in May 2024 at $175 million, upsized 40% from a $125 million target, with the risk margin fixed at 12%, the top of 11% to 12% guidance (Artemis Deal Directory). The World Bank's transaction summary gives the annual risk statistics: expected loss 4.09%, attachment probability 6.26%, exhaustion probability 2.54%. The margin is 2.93 times expected loss. Expected loss over attachment probability gives a 65% conditional severity, and exhaustion over attachment says 41% of triggering storms take the full $175 million. The structure assumed that a storm crossing the line at 937mb or lower usually crosses deep.

Loconte, an ILS investment specialist at Azimut Switzerland, ran 1,000 members of Google WeatherLab's r2 ensemble as Polo stalled and found 77 that triggered the bond, with losses from 25% to 100% and a mean near 45% (Artemis, September 2026; Bloomberg via Claims Journal, September 2026). That is about 3.5% of principal in expected loss from a single storm, close to the tranche's full annual 4.09%. The next run cut the count to 22 with a 75% maximum and the same 45% mean, about 1% of principal. The 965mb crossing took it to zero.

Point on the path	NHC central pressure	Triggering paths of 1,000	Mean loss given trigger	Implied loss of principal
Sep 22, first ensemble run	Deepening toward 892mb	77	About 45%	About 3.5%
Sep 22, second run	926mb, Category 5, offshore	22	About 45%, 75% maximum	About 1.0%
Sep 25, re-intensification	908mb, 180 mph	Track offshore of the gate	n/a	Bid/ask widened
Sep 28, landfall near Las Barrancas	965mb, 110 mph	0	n/a	0%

A fund marking the position moved along that path. Artemis reports bid/ask spreads widened on the uncertainty over track and intensity, without the response a high perceived risk would produce (Artemis, September 2026). Florian Steiger of Icosa Investments said mid-week: "I'd never say never. But right now, I'm not too concerned" (Bloomberg, September 25, 2026). The 45% mean loss across triggering paths sits below the 65% conditional severity in the modelled statistics, because most of the ensemble's crossings were glancing ones at the 25% and 50% steps.

Mexico pays $21 million a year for this tranche, $84 million over the term. The Atlantic Class C in the same program carries 13.5% on a 5.69% expected loss, 2.37 times. The 2020 Pacific Class D carried a 4.06% expected loss at a 6.5% margin, 1.6 times, and it paid the $60 million on Otis (World Bank, May 2024). Same modelled risk, 85% more margin, bought in the 2024 hard market after a payout.

Pricing has since turned: Leadenhall's retro bond priced at 1.67 times, debut sponsors near 2 times, and secondary spreads at 5.71% in late June. Polo landed on a tranche carrying more margin than the sponsor could buy today.

Three Near Misses, the $575 Million Layer Beneath, and the April 2028 Renewal

Since settlement the program's record runs Erick in June 2025, the M7.3 quake in July 2026, and Polo in September. Zero payouts against about $54.6 million a year of program margin at 4% to 13.5% across the four tranches. A Pacific tranche that pays nothing across its first 29 months is the base case at a 6.26% attachment probability. Three near misses in 15 months is what that base case looks like from the sponsor's chair.

Beneath the bond sits the layer Polo may have reached. Mexico doubled its Agroasemex-placed parametric catastrophe insurance to 10 billion pesos, about $575 million, for June 5, 2026 to May 5, 2027, covering earthquake, volcanic eruption, hurricane and flood (Artemis, June 2026). Artemis understands that arrangement carries lower thresholds than the bond, and whether Polo activated it was undetermined at landfall.

AM Best's Alfonso Novelo expected losses concentrated in commercial infrastructure and business interruption from prolonged power outages, in a northern coastal region of low insurance penetration where insurers cede more than three-fourths of local exposure through excess-of-loss cover (AM Best via The Insurer, September 2026). The bond sits above a layer, and the layer sits above a market that mostly does not insure the coast Polo hit.

The complication for the April 2028 renewal is that neither near miss can be fixed without buying the other. Lower the pressure threshold to 950mb and Erick pays; Polo still crosses at 965mb and pays nothing. Push the gate seaward to catch a Category 5 stalled 215 miles offshore and the bond pays on a storm doing no damage, the false positive the 2024 redesign spent its basis-risk budget to avoid.

The ensemble showed the third quadrant: 77 of 1,000 paths where Polo turned into the Guerrero coast at Category 5, where a full $175 million payout would have landed against a loss on the scale of Otis, which the NHC put at $12 to 16 billion (NHC, March 2024). The gate can be moved. Moving it to catch both storms is the one option the geometry does not offer, and the 12% margin is the price of choosing which one.

Hurricane Polo Hit 892mb and Paid Mexico's Cat Bond Nothing

Key Takeaways

The Gate: Pressure and Position Read at the Same Instant

Pricing the Week: 12% Against a 4.09% Expected Loss

Three Near Misses, the $575 Million Layer Beneath, and the April 2028 Renewal

Further Reading

Sources