Skip to main content
main content, press tab to continue
Article | WTW Research Network Newsletter

Volcanic risk and insurance: The untapped potential of downward counterfactual perspectives

By Gordon Woo and James Dalziel | November 30, 2023

Volcanic risks are less frequent than other natural disasters, but their unpredictability poses a unique challenge. Should insurers incorporate downward counterfactual analysis into risk management frameworks to better prepare for these rare but catastrophic events?
Climate|Willis Research Network
Climate Risk and Resilience

Volcanic eruptions are relatively uncommon when compared to the frequency of other geohazards such as earthquakes, but volcanic unrest is a phenomenon that occurs much more often. A period of volcanic unrest is signified by the observation of precursors such as seismicity, ground deformation, release of volcanic gases, and increased heat flow. Periods of unrest can be prolonged and can last for months or even years. For example, in the 1930s there were several years of unrest on the volcanic Caribbean Island of Montserrat. A leading volcanologist, Frank Perret, surmised that if all the energy associated with the unrest were released in an eruption, it would have destroyed the buildings on the island.

There was no eruption following the unrest in the 1930s, nor in the 1960s when there was further major unrest on Montserrat. But both periods of unrest warned of the potential for an eruption, which eventually began on 18 July 1995. Activity over the following two and half years produced powerful phreatic explosions, ashfall, pyroclastic flows, and mudflows. The eruption episodes claimed a total of 24 lives and destroyed nearly two-thirds of the island, resulting in a surprise loss for the insurance industry.

This sequence of events begs the question: Could Montserrat and the insurance industry have been better prepared? Reflecting on the unrest in the previous decades, could a deeper analysis of "what could have been" — employing downward counterfactual thinking — have spurred more robust preparedness measures?

Aerial image taken from NASA Earth Observatory of Montserrat in 2009
Montserrat in 2009

Source: NASA Earth Observatory.

Tepid demand for volcanic risk modelling

The 1990s marked a significant era in the evolution of risk assessment for natural catastrophes. Costly disasters such as Hurricane Andrew (1992) and the Northridge earthquake (1994) highlighted the inadequacy of existing risk management practices and catalyzed the emergence of the catastrophe modelling industry[1]. But while models for weather-related and seismic risks quickly gained traction, volcanic risk models did not enjoy comparable adoption, even in the wake of the Montserrat eruption.

This lack of adoption is not because we are devoid of appropriate modelling techniques. Probabilistic Volcanic Hazard Analysis (PVHA), which utilizes statistical and probabilistic methods to evaluate the likelihood of volcanic hazards within a defined time frame, is a well-established academic research field[2]. However, the rarity of volcanic events and the absence of significant historical losses have led to a tepid demand for such models in the insurance sector, especially when compared to the more frequently occurring perils such as wildfires, floods, and windstorms.

The absence of volcanic catastrophe models means that many insurers may not fully recognize the eruption-related risks to which they are exposed, which include direct physical damage as well as indirect effects such as supply chain disruption. Given this gap, there's an evident need for insurance underwriters and risk managers to consider alternative approaches, such as downward counterfactual analysis.

Reimagining the past to prepare for the future

Downward counterfactual analysis is a risk assessment technique that contemplates scenarios more severe than those that have historically been recorded. This "what if" methodology probes the outcomes of near-miss incidents, assuming they had escalated adversely. The aim is to harness insights from these hypothetical worst-case situations, to enhance preparedness and reduce the impact of similar future events. In 2017, Lloyd’s and RMS published a report on the benefits of downward counterfactual thinking for the insurance industry[3].

For a given historical period of volcanic unrest, it is possible to reimagine the past by considering a range of plausible eruptive scenarios and estimating their likelihood of being realised, however, such analyses could have been undertaken for Montserrat during the periods of unrest in the 1930s and 1960s; a 1971 study by the University of the West Indies[4] concluded that there was an abnormally high risk of an eruption in the near future. However, risk managers and insurers did not take notice of this, and were therefore unprepared when the eruption finally occurred in 1995.

Currently, there is unrest at the Campi Flegrei caldera near Naples in the form of ground deformation ongoing since 2004, and more recent seismic activity. An Mw4.2 earthquake occurred on 27 September 2023, the largest in forty years, which should prompt a reanalysis of past unrest to help prepare for possible future outcomes. Between 1982 and 1984 there was a volcanic crisis at Campi Flegrei, which caused building damage and triggered the evacuation of tens of thousands of people. A downward counterfactual analysis of this crisis could consider the building damage consequences of future seismicity clustered under different towns within the caldera. A study of this kind was carried out in 2020[5], although the downward counterfactual terminology was not used.

Unrest at Campi Flegrei, a densely populated area, poses a significant risk to life and infrastructure in the event of eruption, but impact on the insurance industry would likely be limited due to low insurance take-up in Italy[6]. This situation is replicated in many cities on volcanoes around the world, however, the United States, Japan and New Zealand are countries where there are areas of high insurance exposure that are subject to significant volcano risk. In the past, brokers and model vendors such as RMS have undertaken property portfolio volcano risk analyses for a number of high-risk areas such as the Pacific Northwest and Hawaii. These analyses predated the 2017 Lloyd’s report[5], otherwise downward counterfactuals may have been considered.

Also predating Lloyd’s report was the eruption of Eyjafjallajökull, Iceland in 2010, which caused enormous disruption to European air travel, and temporary closure of European airspace. The International Air Transport Association stated that the total loss to the airline industry from the eruption was around $1.7 billion[7]. When rare events such as this occur, thoughts turn to how such losses might have been mitigated. These are upward counterfactuals, but risk managers need also to be mindful of downward counterfactuals, such as the triggering of an eruption of Katla, the big sister volcano of Eyjafjallajökull. Cascading events, where one event triggers another, can be very difficult for risk modelers to anticipate but devastating if they do occur. Identifying such compound scenarios for historical events is an objective of downward counterfactual analysis.

At the time of writing, there is also ongoing unrest on the Reykjanes peninsula in Iceland. The Fagradalsfjall volcanic system returned to life in 2021 having been dormant for 800 years, with eruptions taking place in 2021, 2022, and earlier this year. The latest earthquake swarms indicate a larger magma volume compared to previous episodes. Currently, the intrinsically stochastic nature of the underlying volcanic dynamics is reflected in the various ways this unrest could evolve. An explosive eruption on the scale of Eyjafjallajökull is highly unlikely due to magma composition and lack of ice cover. However, part of the volcanic system is located offshore, and so downward counterfactuals of the recent unrest could consider implications of an explosive submarine eruption, including the effects of ash on the nearby Keflavik International Airport, or flourine gas production impacting livestock and agriculture (similar to the Laki 1783 eruption).

Using volcano monitoring data to create eruption scenarios

Although the majority of the world's volcanoes are not actively monitored due to a combination of high costs and physical inaccessibility[8], ground-based geophysical monitoring is often in place for volcanoes in developed countries and satellite monitoring of global volcanic ground deformation is becoming an increasingly useful tool. The Smithsonian Institution[9] tracks weekly volcanic activity, and institutions like the Centre for the Observation and Monitoring of Volcanoes, Earthquakes and Tectonics (COMET) use satellite imagery and Interferometric Synthetic Aperture Radar (InSAR) to monitor deformation signals[10]. As volcanic unrest can precede an eruption, some observed periods of unrest may be considered as eruption “near-misses” and could be used to create downward counterfactual scenarios of the loss consequences of the unrest evolving into an eruption.

Incorporating downward counterfactual analysis into risk management practices is not just an academic exercise, but an essential strategy for ensuring financial robustness and strategic foresight. By reimagining worse outcomes than those historically observed, insurers can anticipate and prepare for events that, while rare, have the potential to cause catastrophic loss. Recognizing these near misses allows for more refined pricing, customized coverage options, and adequate capitalization. As the world faces increasing uncertainty, integrating downward counterfactual analysis for volcanic risks ensures that the insurance sector remains prepared, adaptable, and valuable in its role as a safeguard against the unexpected.


  1. Jones, M., Mitchell-Wallace, K., Foote, M. & Hillier, J. Fundamentals. In: Mitchell-Wallace, K. et al. (ed) Natural catastrophe risk management and modelling 1–46 (John Wiley & Sons, Ltd, 2017). Return to article
  2. Connor C, Bebbington M, Marzocchi W. Probabilistic volcanic hazard assessment. In: Sigurdsson, H. (ed) The encyclopedia of volcanoes, 2nd edn. (Academic Press, 2015). Return to article
  3. Woo G., Maynard T., Seria J. Reimagining history: counterfactual risk analysis. Reimagining history report. (2017). Return to article
  4. Shepherd, J. B., Tomblin, J. F. & Woo, D. A. Volcano-seismic crisis in Montserrat, West Indies, 1966–67. Bull Volcanol 35, 143–162. (1971). Return to article
  5. Charlton, D., Kilburn, C. & Edwards, S. Volcanic unrest scenarios and impact assessment at Campi Flegrei caldera, Southern Italy. Journal of Applied Volcanology 9, 7. (2020). Return to article
  6. Associazione Nazionale fra le Imprese Assicuratrici. Italian Insurance 2021-2022. (2022). Return to article
  7. Ash chaos ‘cost airlines $1.7bn’. business. (2010). Return to article
  8. Sparks, R.S.J., Biggs, J., & Neuberg, J. W. Monitoring Volcanoes. Science 335, 6074. (2012). Return to article
  9. Global Volcanism Program | Smithsonian / USGS Weekly Volcanic Activity Report. (2023). Return to article
  10. Anantrasirichai, N., Biggs, J., Albino, F. & Bull, D. A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets. Remote Sensing of Environment 230, 111179. (2019). Return to article

Catastrophist, Moody's RMS

Head of Earth Risks Research
email Email

Contact us