The Big One Two

A 7.6 Mw quake shook Japan on the 8th and caused a two-foot tsunami wave. Multiple aftershocks followed. Local authorities warned of heightened potential for a “megaquake.” They were worried by mounting evidence that one sizeable earthquake can trigger another, nearby seismic event along another fault line. The chance, they said, was only 1%, but it was enough to prompt mass evacuations.

Fortunately, it was all relatively minor in the end. Yet the quake and subsequent warning should raise the reinsurance industry’s interest in the potential for stress-triggered pairs or clusters of earthquakes in seismically active zones, including along the risky US west coast.

An “earthquake-triggered mainshock” occurs when one major earthquake triggers another massive seismic event elsewhere. This phenomenon has now been shown to have occurred in various places around the world, going back millennia. Everyone is familiar with earthquake aftershocks, but these are different. The follow-up quake isn’t an aftershock, but a serious event in its own right. The epicenter will be in a completely different place, although some localities may suffer back-to-back shaking events. That means buildings weakened by the first, triggering quake could be collapsed by the second, even greater rupture of a different fault.

One possible double-whammy scenario is an earthquake in the Cascadia Subjunction Zone (which runs offshore past Vancouver, Washington, and Oregon) triggering a big rupture of the San Andreas Fault (which runs from Eureka past San Francisco). That raises the specter of a one-two punch hitting the insurance industry: two enormous earthquake losses in short succession.

In an article earlier this year in Geosphere, Chris Goldfinger, a Professor Emeritus at Oregon State, and colleagues reported that “northern San Andreas Fault [earthquake] events have triggered turbidity currents in the southernmost Cascadia Subduction Zone, and vice versa.”

In plain English, the Earth holds evidence of exactly the one-two punch described. Goldfinger and his team of paleoseismologists study the geological footprints of earthquakes that happened thousands of years ago. It looks to them like Cascadia quakes have triggered major earthquakes along the San Andreas, and could do so again. Such paired earthquakes could occur, the researchers say, with just “minutes to hours of separation.”

But we don’t need to go back 3,000 years to see trigger-linked earthquakes. Several recent seismic events seem to illustrate triggering. The massive quake in Turkey in 2023, for example, was an earthquake-triggered mainshock. It was the second major quake of a “multi-fault rupture” that began with a trigger event, the 7.8 Mw earthquake along the East Anatolian fault nine hours earlier. That quake seems to have led the Mw 7.7 earthquake on the North Anatolian fault which caused the insurance sector’s largest earthquake loss of the decade so far.

According to scenarios from Moodys RMS, a repeat today of the 1700 Cascadia earthquake, which had an estimated magnitude of up to 9.2 and caused tsunami waves in Japan, would produce an insured loss of around $2 billion. A repeat of the infamous 1906 San Francisco earthquake on the San Andreas fault would cost insurers around $130 billion.

The Cascadia loss estimate is low, for multiple reasons. It includes losses only from shake and fire-following, but it’s the wave that would cause the most extensive and costly damage. Research in 2015 by the USGS found the tsunami caused by Cascadia 1700 could cause total damage of $70 billion. It would wipe out 100 bridges, take out Route 101, and kill as many as 10,000 people.

Although private flood insurance penetration is rising in the US, it remains a relative rarity, and most insured flood losses in the US fall to the state-backed NFIP, which is already more than $20 billion in debt. One modelled estimate from about a decade ago puts insured losses in Canada from a 1700-style Cascadia Subduction Zone event at C$18 billion to C$26 billion, or up to about US$19 billion, but inflation will have increased these numbers dramatically since then.

In addition, the modelled losses from both earthquakes exclude classes such as business interruption, accident and health, travel, marine classes, and many others. And damage in Japan is excluded from the estimate. These would add billions at least to a one-two Cascadia 1700 and San Fran 1906 punch.

Putting this all together, we end up with an insured loss well in excess of $150 billion. The economic losses would of course be much higher. The human misery would be tremendous, since relief access to devastated areas would have been cut off by the first event.

A number of modelling firms have already accounted for earthquake-triggered mainshocks in some or all of their models. The just-released Moody’s RMS Europe Earthquake HD Model has added “complex multi-fault ruptures” to its scenarios, the company says, although it has yet to update its US model. In the interim, Cotality (formerly CoreLogic-Eqecat) has incorporated the possibility of earthquake-triggered mainshocks into its US model.

Verisk (formerly Nasdaq Risk Modelling for Catastrophes) and Moody’s RMS are very likely to include clustering when they release US model updates in the years ahead. In the meantime, Willis Re is working with the Willis Research Network to develop clustering adjustments for the RMS and Verisk models which will form part of the Willis Re view of risk.

Major model changes often cause an industry hiccup. RMS v11, a major update to the modeler’s US hurricane tool in 2011, resulted in large loss-estimate increases that caught many by surprise, because prior model updates had been more modest. The release had a large impact on reinsurance pricing, and led rating agencies to downgrade bonds that had been structured on the previous model’s loss assumptions.

Events sometimes prompt upgrades that move markets. The 2011 Tohoku earthquake reached Mw 9, shattering the widely held belief that events of such great magnitude were not possible on subduction zone faults. New models released after 2011 included them, which increased earthquake tail risk. On the wind side, after the busy 2005 hurricane season caused large losses, RMS and Verisk introduced new event sets that account for near-term climate conditions. These alternative views of hurricane risk were deployed after the market raised concerns around the potential impacts of climate change.

A current market rumble has followed recent model updates both by RMS and by Verisk for US Severe Convective Storm risk. Modeled pricing has increased with the updates by 50%, and in some cases 100%. The changes reflect large increases in actual SCS-driven claims over the past five years. As a result, many reinsurance markets have scaled back their SCS capacity. Market response to expected updates of the pair’s US earthquake models to include triggered earthquakes cannot be predicted until their scope is known. However, they have the potential to shake up thinking around some of the sector’s most serious loss scenarios.

SCS and other “secondary” perils may be the ones that have reinsurers double-checking their exposure lately, but earthquake risk should not be disregarded as mundane. It is widely recognized as one of the reinsurance market’s top-two peak catastrophe exposures, alongside Florida wind. It is a leading driver of hedging strategies, and coverage is widely available, particularly under Excess of Loss Occurrence and Aggregate treaties, under quota shares, and under indexed and parametric structures.

The 2023 earthquake-triggered mainshock in Turkey was the biggest earthquake to impact the reinsurance market in recent years, but despite coming close for some carriers, it didn’t impact retro programs. A one-two punch from Cascadia-San Andreas would be a very different case indeed.