From peril to system: 15 years after Tohoku, are we prepared for the extreme?

Pavel Huerta

Head property underwriting globals for APAC, Swiss Re

On 11 March 2011, the ground moved. So did our assumptions.

The Great East Japan Earthquake and ensuing Tsunami were first and foremost a human tragedy, claiming 18,000 lives and inundating more than 500 km of coastline. For the insurance industry, it was also a structural stress test. It forced us to look beyond vulnerability curves and ask a harder question: what happens when a complex system fails?

15 years on, our technical capabilities are far more sophisticated. Tsunami is no longer an afterthought. Catastrophe models now account for the possibility that several connected parts of a major fault could break at once, even off the Tohoku coast, leading to powerful shaking and tsunamis.

Monitoring networks are stronger. Hazard envelopes are broader. Vulnerability assumptions are now calibrated on far richer post-event data.

The models have improved.

But the most important lesson from Tohoku was not about tsunami wave height. It was about uncertainty.

The scale of the shock — and the role of reinsurance

The 2011 event generated approximately US$35bn in insured losses and around US$210bn in total economic losses, making it the costliest natural catastrophe globally at the time. Despite Japan’s stringent building standards limiting structural collapse, 2011 became a clear peak loss year for the global reinsurance industry.

For Swiss Re, the net impact from the earthquake and tsunami alone was approximately US$1.2bn. Despite the magnitude of the event, claims were paid, primary insurers remained solvent, and capacity was available at the following renewals. What began as a domestic catastrophe became a globally diversified financial event.

These figures are more than statistics. They demonstrate how internationally diversified reinsurance capital absorbs severe losses, stabilises cedents during acute uncertainty, and preserves market function after systemic shocks.

Through multiple peak years, the reinsurance industry has remained a consistent provider of capacity — not only in favourable markets, but when capital is most needed.

That continuity is not incidental. It is the core function of disciplined, globally diversified reinsurance.

The Illusion of Precision

Catastrophe models are indispensable. They quantify tail risk. They allow us to structure programmes and optimise capital. They allow us to negotiate coverage with discipline.

But they are not reality. They are structured representations of what we understand today.

Tohoku and other peak loss years do not invalidate modelling, they test its boundaries. The Great East Japan Earthquake demonstrated that even in one of the world’s most earthquake-prepared nations, hazard assumptions can be exceeded. They also underscore the need for stress-testing beyond consensus scenarios.

At the time, many experts and officials did not anticipate an earthquake of such magnitude on the Japan Trench off Tohoku. The largest magnitude that was thought possible was in the low-8 range. Coastal defences, designed for one level of protection, were overtopped. A nuclear incident at the Daiichi Power Plant reshaped energy policy. Semiconductor plants thousands of kilometres away suddenly mattered to insured losses.

Isolated Perils in a Connected World

Our industry is organised around peril categories: earthquake, wind, flood, and wildfire. Even when we look at secondary perils, we usually remain within a single hazard family.

The world does not operate that way.

Today’s exposure base is highly geographically concentrated, operationally interconnected and increasingly dependent on digital infrastructure. Power, telecoms, transport and water fail together. Supply chains are optimised for efficiency, not resilience. Financial markets amplify volatility, and social behaviour can also alter loss outcomes.

The critical question after a major event is no longer “How large was the earthquake?” It is “How did the system respond?”

That distinction matters when allocating capital at the extreme.

Uncertainty Is Structural

Every major catastrophe in the last two decades has revealed a blind spot.

• Tohoku highlighted tsunami underestimation and infrastructure interdependency.
• Christchurch, which caused insured losses of ~US$30bn exposed the scale of liquefaction risk, significantly increasing residential and infrastructure claims.
• The Thailand floods, which caused insured losses of ~US$20bn, revealed global supply chain concentration. Major impacts on electronics, automotive and hard drive manufacturing caused global business interruption far beyond Thailand.
• COVID, which caused estimated insured losses of ~US40-50bn+ globally, demonstrated non-physical business interruption exposure.
• Recent extreme weather events have shown how hazard clustering challenges independence assumptions.

These were not failures of the modelling. They were reminders that models operate within defined boundaries. Reality does not.

The next surprise is already somewhere in today’s portfolio. We simply do not yet know its form.

That is not cause for alarm. It is a reminder of what reinsurance is designed to do and what it should not do.

Capital Is for the Unknown

Reinsurance exists to absorb volatility beyond the expected. Primary insurers manage frequency and moderate severity within diversified books.

Reinsurers sit at the edge. We provide capital where uncertainty is greatest.

If tail risk were fully known and stable, it would not require reinsurance. The tail remains long because uncertainty remains structural and has implications for how we think about:

• Return period stability
• Correlation across regions and lines
• Event independence
• Aggregate management in a world of compound risk

Capital models rely on structured scenarios. Capital buffers must also recognise that the next systemic interaction may not resemble the last one.

The cost of capital at the extreme is therefore not just a function of expected loss. It is a function of uncertainty around that loss.

Managing the extreme, not the Mean

In a competitive environment, it is tempting to anchor on recent losses and model averages. But catastrophe risk is defined by extreme, peak years.

The real test is what happens in the tail. That means looking beyond expected losses and asking harder questions: how much capital is exposed in a truly severe scenario? How much uncertainty sits around the model? Where might our assumptions be wrong?

A 1-in-200-year loss is not a precise number — it is an estimate in a world that keeps changing.

Ignoring uncertainty around severity, interconnectivity and how fragile interconnected systems can become, or how losses can spread beyond the original footprint, runs the risk of underestimating the exposure we are taking on.

This is not theoretical. It is fundamentally about protecting capital so that we can continue to be there when clients need us most.

A business imperative – from lessons learned to resilience

The question is not whether models are useful. They are. The question is how much confidence we assign to their precision when allocating capital at the extreme edge.

15 years after Tohoku, our view of tsunami risk is stronger. Datasets are richer. Vulnerability modelling is more robust. The event also prompted further strengthening of Japan’s already high disaster mitigation standards, including incorporation of very low-frequency, high-impact scenarios into national planning. 

Yet Japan remains one of the most earthquake-prone countries in the world, and the potential for another peak seismic loss year is ever-present.

Reinsurers have an important role to play in preparing for what comes next. By working closely with insurers, corporates and governments, we can help design solutions that expand earthquake coverage and gradually narrow protection gaps over time.

Taken together, this is about combining capital, expertise and partnership to strengthen resilience over the long term.

The lessons from the Great East Japan Earthquake remain highly relevant today. Hazard models must constantly evolve, mitigation must strengthen, and risk transfer mechanisms must scale with exposure. Peak loss years are not anomalies; they are structural features of catastrophe risk.

Reinsurance exists to absorb shocks, to enable recovery and to sustain confidence when uncertainty is greatest. The Great East Japan Earthquake was a defining test of that purpose. The responsibility now is to ensure that when the next peak earthquake arrives, the system – and society – is even more resilient.

The next major loss will not look like 2011. It never does.

Our responsibility as reinsurers is to ensure that discipline, structure and capital allocation reflect that reality.

That is not pessimism. It is prudence. And it is the foundation of a resilient reinsurance market.