Insurance as Adaptive Capacity
Think back to when you last travelled to someplace new. Did you verify the building you’d be in was structurally safe and up to code? Did you learn which fire department you should call in case of emergency? Do you even know precisely which fire district you are located in at this very moment?
Chances are: you didn’t, you didn’t, and you don’t. And for good reason. We simply take it for granted that buildings we visit comply with sound building codes, and we take it for granted that fire departments offer emergency responders, always just a phone call away. This near-universal coverage in much of the developed world is offered and maintained through government, but both measures trace their history to the insurance industry.
The decades following the Great Fire of London in 1666 saw the earliest development of private fire brigades run by insurers, initially focused on their own policyholders, at a time when “there were no such things as organized municipal fire protection or fire-fighting appliances worthy of the name” (Johnson 1972). These companies affixed a firemark to each insured property, which signaled a fully paid premium and eligibility for fire department services (see example in Figure 1). In time this practice yielded to government-run fire departments. The Great Fire and other historic fires also led to the development of modern fire safety and building codes, with substantial leadership and interest from property insurers of the time.
In climate science, adaptation refers to the process of reducing or eliminating the adverse consequences of climate change (this is distinct from mitigation, which is reducing or eliminating the climate change itself). Adaptive capacity is our society’s overall ability to adjust to climate change consequences. Evan Mills (2005) succinctly stated “insurance is a form of adaptive capacity.” Through actuarially sound risk pricing and underwriting standards, insurance markets nudge communities towards risk-reducing behaviors, and help societies adapt and grow more resilient in an uncertain world.
This extends to climate and weather risk. A useful statistical distinction between climate and weather is to consider climate the distribution of weather. That is, weather events form individual outcomes, and climate describes the range of possible outcomes and their relative frequencies. Climate change is a shift in the distribution of weather. From these definitions, one can see that weather risk is the possibility of an adverse weather outcome such as a hurricane, drought, etc.; climate risk is an overall shift in the distribution of these events, making them more likely or more intense.
Insurers seek to measure both climate and weather as drivers of loss. An example of the former is incorporating loss experience into premiums for flood insurance. “In a perfectly efficient and well-informed insurance market, premiums for flood-risk cover should be determined by the risk of flooding, which is a property of the climate, not the actual weather” (Allen 2003). Huq, Roberts, and Fenton (2013) described the role of insurance in addressing “unavoidable” losses and damages from climate change. However, they did not specifically comment on its potential for assisting with “avoidable” loss and damage, which would include losses averted through climate change mitigation, or by enforcing risk reduction measures as part of its underwriting process. One can ask if there is untapped potential for insurance to nudge climate adaptation along further: “can insurers extend their self-chosen historical role in addressing root causes (as founders of the first fire departments, building codes and auto safety testing protocols) to one of preventing losses at a much larger scale, namely, the global climate?” (Mills 2005). Given the emphasis on the global scale, one could ask the same of reinsurers.
Developments Since 2005
This question was asked in 2005. Despite much progress and strong, continued interest from both insurers and reinsurers, it is still the right question to ask today.
The highest cost weather events are having an increasingly large impact on total losses. The U.S. National Oceanic and Atmospheric Administration’s National Centers for Environmental Information (NOAA NCEI) boasts 37 petabytes of climate data, focusing on both the U.S. as well as the entire globe. This is the largest repository of environmental data in the world. Among these data is the billion dollar disasters database, which tracks all weather-related losses in the United States whose total costs exceed $1 billion U.S. (NOAA 2019). These large loss events comprise an increasingly large share to total weather-related losses, moving from roughly 75% of total weather-related loss costs over 1980-2000 to roughly 85% of costs over 2000-2018 (Smith 2019). Swiss Re (2020) reports globally that both the number and total dollar loss amounts from weather-related events are increasing. Since reinsurance is generally triggered after insured losses exceed a high threshold, this shift in the loss cost distribution curve indicates a shift in liability exposure towards the reinsurance industry. The Intergovernmental Panel on Climate Change assessment reports project an increase in weather extremes (IPCC 2014), suggesting more of the above for the reinsurance industry.
This shift is illustrated very clearly in Emanuel (2017), in which the rainfall produced by Hurricane Harvey in Texas was recast under both past and future climate scenarios. The hurricane was estimated by NOAA’s Office for Coastal Management to have a total cost of $125 billion, and the Insurance Information Institute lists it as the fifth costliest hurricane for insured property losses at estimates of $16-19 billion. It is safe to say that Hurricane Harvey is of tremendous interest in the insurance and reinsurance communities. Focusing only on climate and holding other social and economic factors constant, Emanuel (2017) found that the associated rainfall had an annual rainfall exceedance probability of around 1% under the 1981-2000 climate; that number rises to 18% under the 2081-2100 climate described by the RCP 8.5 scenario (IPCC 2014). This means the “once-in-one-hundred-years” rainfall event of Hurricane Harvey could be roughly a “twice-a-decade” rainfall event under this particularly aggressive climate change scenario.
Swiss Re conducted a similar retrospective study by asking what might happen if Hurricane Katrina hit New Orleans in the year 2020, instead of 2005. In 2005, the event led to an estimated $160 billion (2020 dollars), of which $41 billion was privately insured. Pourrabbani (2020) estimated that despite population decrease in the city, the shift in exposures could lead an identical storm to cause $175 billion if it had hit in 2020, with $60 billion in privately insured losses. They went on to estimate that increasing the wind speed by just 5% would increase the storm surge and add another $5 billion to insured losses and bring the total loss above $200 billion. Given that the region is experiencing rapid sea level rise, the study concludes that an identical hurricane to Katrina could achieve this higher storm surge level in perhaps as little as a decade.
New products, research committees, and insurance industry priorities around climate risk have emerged in the last fifteen years. Brown, Osgood, and Carriquiry (2011) describe many “science-based” innovations in insurance, such as index-insurance and the use of remotely sensed data products. The 12th Annual Survey of Emerging Risks – jointly run by the Society of Actuaries (SOA), Casualty Actuarial Society (CAS), and Canadian Institute of Actuaries (CIA) – listed climate change as the top risk of 2019 (Rudolph 2019). The SOA chose “catastrophes and climate” as one of five key strategic research programs in its 2017-2021 strategic plan (SOA 2019) and formed its Climate and Environmental Research Sustainability Committee in 2016, and the CAS formed its CAS Climate Change Committee in 2009. These build on the momentum of individual insurance and reinsurance companies articulating their commitment to managing climate risks (Mills 2009, 2012).
Looking Ahead
Surminski, Bouwer, and Linnerooth-Bayer (2016) issue warnings on two fronts. First, that weather-related property insurance could transfer costs to the most vulnerable populations at the forefront of risk. For example, residents of floodplains can be disproportionately lower income, and so asking these populations to fund the insurance to protect their rising risk could merely shift costs to the most vulnerable. Second, subsidizing insurance premiums could undermine the incentive for risk reduction and the “adaptive capacity” function of insurance markets. As an example, agriculture insurance subsidies can incentivize farmers to plant unsuitable crops in certain areas, or assume more risk than they otherwise would (Hazell, Sberro-Kessler, and Varangis 2017).
To further explore these issues in an interdisciplinary setting, researchers at the Statistics and Applied Mathematical Sciences Institute (SAMSI) in the United States and the Reinsurance Association of America (RAA) teamed up to host a workshop The Nexus of Climate Data, Insurance, and Adaptive Capacity in Asheville, NC in November 2018 (Erhardt et al. 2019). Fifty-seven researchers from climate science, government, and private industry met to discuss the issue from their respective points of view. Topics of discussion included the use of climate model projections in insurance and reinsurance work, rethinking flood risk beyond existing flood maps and return levels, ways in which insurance and reinsurance companies can directly stimulate mitigation and adaptation efforts, and broad discussion of extreme events framed as “what keeps executives at insurance companies awake at night?” The full set of outcomes can be found in Erhardt et al. (2019), but a few themes emerged and highlighted why insurance hasn’t met its full potential for climate adaptation. First is the need to move beyond the binary “in-or-out” measure for flood zones, and towards more nuanced maps which communicate risk and uncertainty across a range of values; existing maps and flood insurance do not properly convey risk and shape development patterns. Next, climate models are often run at spatial or temporal scales which are of limited use for insurance and reinsurance. Insurers need to see more research at the decadal time scale, and greater confidence that climate models are producing reliable projections at the regional scale. The shift towards extreme events driving insurance losses was noted, but insurers spoke of the need to understand how individually non-extreme causes can combine to produce an extreme insurance outcome. For instance, strong but not overly extreme rainfall coupled with a rapid rise in temperatures before the ground had fully thawed occurred in the Midwestern U.S. in early 2019, causing the intense flooding at that time. No individual contribution was particularly “extreme” but the combined effect was extreme.
Looking back at fire departments and building codes, with the benefit of hindsight one can see why insurers were so successful. There were clear and affordable measures to reduce individual losses, both the costs and benefits are easily attributable to the individual policyholder, and policyholders and insurers had aligned incentives. With climate, this is often not the case. Vajjhala and Rhodes (2019) write that many resilience measures such as seawalls have distributed costs and benefits across many policyholders, benefits will accrue much later in time, and individual insurers do not have aligned incentives to encourage investment in risk reduction today for distributed benefits that may not arrive for decades. To remedy this, Vajjhala and Rhodes describe how insurance-linked finance stands out as a promising path forward for actuaries and insurers. Large-scale projects which would reduce climate losses over a large number of policyholders and over a long time period could be designed with actuarial and risk modeling to estimate these insurance savings. There are a number of ways to require insurance among the property owners – mandating property insurance within the specified area, imposing an assessment, and so forth. The financing of the project could explicitly draw on the expected insurance savings over the long term, making the project more appealing to those needed to contribute capital in advance. The entire idea requires actuarial expertise, and thus represents a strong opportunity for practicing actuaries.
The fifteen years since Mills called insurance a form of adaptive capacity have seen advances in measuring and managing extreme weather risk, and a strong adoption of the consequences of managing climate risk by insurers and reinsurers. The next fifteen years will require greater research collaboration between climate scientists and the insurance industry to fully realize the untapped potential for insurance markets to aid in climate adaptation. This partnership would enable the best science-driven estimates of climate risk and weather-related premiums needed to address that risk. Allen (2003) wrote of insurers measuring climate risk that “Crucially, however, the size of the ‘climate-change risk premium’ would be determined by the hidden hand of the market, not by politicians in tortuous intergovernmental negotiations. There would no longer be any need to forge a near-global consensus on the risks of climate change before we agree on what to do about it.”