Ian revisited: Disentangling the drivers of US hurricane losses

Even adjusting for inflation, the September 2022 storm, one of history's most-destructive natural catastrophes, would have been half or even a third as expensive in the 1970s, data shows. At its one-year anniversary, a new Swiss Re analysis pivots off the dynamic drivers behind Hurricane Ian's costs and offers insights for the US Southeast, Gulf Coast and Northeast, regions with diverse characteristics but something critical in common: the importance of adapting our built environment to a future of volatile weather. 

When Hurricane Ian took shape off Africa's west coast in mid-September 2022, it was already late in the North Atlantic tropical cyclone season. Before it was finished, however, Ian inflicted about USD 65 billion in insured losses, much of them concentrated near Fort Myers. While Ian's dimensions weren't extraordinary as hurricanes go – it made landfall at Category 4 – several factors combined to make it the third most costly hurricane ever in the US, including a Florida population that's tripled to over 22 million since 1975 and the accompanying concentration of valuable, vulnerable assets in the storm's path. 

Swiss Re's Natural Catastrophe models offer the opportunity to quantify individual drivers of losses associated with hurricanes like Ian, to understand the interplay of multiple factors that conspire to push costs higher. Our modeled expected losses take into account mitigating measures like updated building standards, while also considering variables including the potential contribution of climate change. 

This analysis focuses on residential exposure and wind-only modeled losses to compare modeled annual expected loss scenarios from 1975 and 2020 for multiple US regions exposed to hurricanes. And while the high cost of rebuilding after Hurricane Ian has been fanned1 by surging post-COVID-19 pandemic price increases, our analysis (in today's USD) excludes impacts of inflation. 

For the broader US, our analysis shows population growth is the most significant driver behind increases in annual expected hurricane losses, more than doubling anticipated losses compared to the 1970s.  Coastal regions that are exposed to a unique set of threats including tropical cyclones have drawn significant numbers of new residents, with US Census Bureau figures showing more than 60 million people live in areas most vulnerable to hurricanes after rapid population growth in recent decades.

Storm frequency also plays a role. The last 150 years of observation show significant variation in North Atlantic hurricane activity, mainly fueled by sea surface temperatures (SST). Typically, warmer SSTs result in more hurricanes. The post-1995 period has been marked by elevated activity. Relative to historically low 1970s hurricane activity, today's higher frequency leads to an increase in expected residential market losses in the US, though the magnitude of such differences depends on the period used for comparison.

While in the past, SSTs fluctuated over a period of multiple decades, higher SSTs are now being influenced increasingly by heating from greenhouse gases, a phenomenon that now prevails as aerosol-induced cooling has eased with improved air quality around the North Atlantic. With greenhouse gas emissions and SSTs likely to continue increasing, a return to lower hurricane activity closer to historical averages is unlikely, although there's no scientific consensus on how climate change will ultimately influence future North Atlantic hurricane numbers or landfall frequency. 

With both population growth and hurricane frequency contributing to loss expectations, they are only partially offset by the vulnerability-reducing impacts of improved US building practices. These improvements mitigate expected loss development by just 20-30%, our data shows. While buildings erected in high-growth regions generally follow more recent codes, overall increases in value concentration overshadow the mitigating influence of adaptation efforts on modeled losses.

Overall, population trends and the concentration of assets in coastal urban areas have been the major drivers for US hurricane loss expectations, Swiss Re data shows. Higher hurricane activity relative to the 1970s is continuing, fueled by warm sea surface temperatures. These drivers more than offset reductions in vulnerability linked to building standards. 

The following chart for the US market (residential only) illustrates the interplay between these factors to show the change since the 1970s.

Fast-growing Florida has seen the largest increase in population-amplifying loss potential, amounting to +180-190%, though improved building standards have mitigated loss potential by around 90-100% compared to the 1970s. Ian showed that recently updated or renovated buildings performed very well during the storm, underscoring the value of adaptation measures.

By contrast, in the US Northeast population growth and related property value increases have been less intense since the 1970s. However, efforts to address vulnerability have also been less pronounced, leading to an expected modeled loss reduction of only 10-15% compared to the 1970s. The region's older infrastructure that has seen less advanced adaptation measures is likely more susceptible to hurricane damage, indicating considerable loss potential despite generally fewer intense events. 

North Atlantic hurricane frequency is presently higher than in the 1970s. The impact of this higher frequency on modeled loss expectations is largely uniform from region to region, at around 20-25%, with only slight differences depending on whether weaker or stronger hurricanes are prevalent. The impact is highly dependent on the observation window.

The drivers of modeled expected losses aren't uniform from region to region. Loss potential in Florida has been most strongly affected by population growth, while population growth has been less pronounced in the Northeast. Improved building standards have had a moderating effect, predominantly in high-hazard, high-growth regions but less so in slower-growth parts of the country. 

The following chart shows the delta by region (residential only). All values are in today's USD.

In Florida's fast-growing Fort Myers area, where Hurricane Ian made landfall, the impact of population growth on modeled annual expected losses has been significant, some 340-350% higher today than the 1970s. Comparatively slower growth in Miami means that population increases there contribute less to modeled loss expectations (160-170%) than they do for the state of Florida as a whole.

Metropolitan and coastal areas in the US Southeast and Gulf Coast states have also seen significant influxes of people and resulting asset concentrations. In Houston, America's second-fastest growing metropolitan area with now more than 7 million people, the population-driven modeled loss since the 1970s has risen by 200-210%. 

In the US Northeast, urban populations have grown more slowly since the 1970s; here, the main period of population and value growth preceded 1950. As a result, there are fewer new buildings that have benefited from improved standards of recent decades. Additionally, given generally weaker and less frequent hurricanes in the Northeast compared to Florida or the Gulf Coast, older buildings have not been the focus of adaptation, leaving them vulnerable. Nonetheless, Hurricane Sandy in 2012 and Hurricane Ida in 2021 have shown the destructive potential of tropical cyclones, leading to increased efforts to strengthen resilience.

Location plus population growth is critical. Coastal metropolitan areas in the South such as Houston or in Florida have seen an intense influx of people, leading to property value concentrations. When a hurricane hits, losses here are likely to be disproportionally higher, with adaptation failing to keep pace with elevated risks from population growth.

The following chart shows the expected loss delta for selected metro areas as well as the corresponding state between 1975 and today for the U.S. market (residential only). All values are in today's USD.

This analysis demonstrates how the capabilities of catastrophe models can be leveraged, not just for pure costing, but also to understand the underlying dynamics of the modeled losses. With the ability to tailor specific elements in the model, we can isolate individual factors that influence modeled losses to capture a feel for their size and contribution to the annual expected loss. Cat models’ capabilities extend much further than this analysis, allowing risk managers, for example, to go beyond the annual expected loss that was the focus of this investigation to capture expected losses from tail events, as well.

The profound impact of population dynamics that emerges from this analysis highlights how important it is to capture current exposure values accurately and stay abreast of potentially rapid developments in insured values, especially in high-hazard coastal and urban regions.

Beyond exposure, population growth also affects how we examine historic losses. In natural catastrophe insurance, it is common to trend losses from historic hurricanes to today’s circumstances, as if these events were to occur today, in an effort to understand how losses develop over time based on numerous dynamic variables. Adjusting historic event losses in “as-if scenarios” that quantify their economic impact were they to occur today cannot be based on merely inflation or GDP growth. The dynamics of property value growth and regional concentration must also be taken into account.

Studies like this operate in a purely modeled world. While this allows us to isolate important loss drivers, their resolution doesn’t reflect the full complexity of the real world. Comparisons of annual hurricane activity and historical landfall frequency depend on the chosen observation window. The expected modeled loss calculation in our analysis is significantly influenced by our choice to compare 1970s hurricane activity with activity today.Despite this sensitivity, however, we can say that we are now in a phase where North Atlantic hurricane activity driven in large part by historically high SSTs has deviated upwards from long-term historical hurricane activity.

It is also important to note that this analysis excludes specific examination of the impact of hurricane-related flooding, a phenomenon which may be exacerbated by sea level rise or increased rainfall. Both rising seas and intense precipitation can increase hurricane losses, not only in coastal regions but also further inland. They are among factors being driven by anthropogenic climate change.

Our models show a clear imbalance: Reducing vulnerability by strengthening building codes, a key element of adaptation, has been insufficient since the 1970s to compensate for expected losses from accompanying population-driven property value growth. Our models highlight a real-world need for adaptation efforts to accelerate, especially in high growth areas but also in areas where growth has been more modest. Moreover, vulnerability-reduction measures also must go beyond wind code improvements to address flood and storm surge protection.

One year after Ian, our analysis illustrates how key macro trends like urban population growth, the concentration of valuable assets in vulnerable regions and lagging adaptation can drive hurricane losses higher. With the average annual growth rate of 5‒7% in insured losses from all natural catastrophes over the past three decades, these trends apply beyond hurricanes, extending to all manner of perils. Understanding the composition and drivers of past losses via natural catastrophe models sharpens our view on how to mitigate future risks and better allows our industry to contribute to overall societal resilience.