Extracts (and data updates) from the article

A re-examination of the U.S. insurance market’s capacity
to pay catastrophe losses

Dionne and Desjardins (2022)

Risk Management and Insurance Review 25, 515-549
Open access

31 July 2024
​
​

• ​Climate finance

Climate finance is defined by the United Nations Framework Convention on Climate Change (UNFCCC) as “local, national, or transnational financing—drawn from public, private, and alternative sources of financing—that seeks to support mitigation and adaptation actions that will address climate change” (reported in Hong et al., 2020). Such financing is intended to change the world economy and build resilience to climate change.

Many financial sectors, ranging from banking and insurance to real estate, are directly impacted by the risks from tornadoes, wildfires, and floods. This raises difficult questions, which were recently discussed in a special issue of the Review of Financial Studies, edited by Hong et al. (2020): How can financial market prices mitigate risks from global warming? How can capital markets raise sufficient financing? How should the distribution of damages from catastrophic events be managed? Despite this type of reflection, however, no studies in finance or insurance have looked at the causal effects of climate change on the insurance industry, though various correlations have been documented.

Here is a typical question in the recent financial literature: Given the potential impact of climate change, are asset prices or firm values sensitive to exposure to climate risks? Three recent contributions address this important question on market efficiency in pricing these risks. Murfin and Spiegel (2020) use information on recent residential real estate transactions to determine whether house prices reflect the differential risks of sea levels rising. They obtain limited house pricing effects with their methodology. By contrast, Baldauf et al. (2020) use transaction data to measure the effect on house prices of flooding projections for individual homes and local measures of beliefs about climate change. They demonstrate that houses projected to be underwater are sold at a discount. Issler et al. (2020) study wildfires in California between 2000 and 2018 with a comprehensive data set that merges information on fires, mortgages, property characteristics, and weather zones. Using the difference-in-differences approach, the authors find a significant causal increase in mortgage delinquency and foreclosure after fire events.

A crucial input in the analysis of climate change risks is the causal impact of climate events on economic activity, which is known as the distribution of damages. It raises an important question about the modeling and sharing of extreme weather risks. Do extreme weather risks, such as the impact of Hurricane Sandy in 2012 or of the 2018 California wildfires, have long-run causal effects on insurance markets? These distributions of damages depend on location-based decisions by households and firms, and technological (self-protection and self-insurance) decisions on preventing and mitigating disaster damages. They also depend on market insurance coverage (including moral-hazard and adverse-selection effects). By modeling these loss distributions suitably, the insurance industry should be able to play a critical role in facilitating risk-sharing and extending insurance coverage for extreme weather events. These research results should also improve public authorities’ role in improving social resilience against climate risk (GAO, 2007; Postal, 2008; Hallegate, 2012, 2014).

 

• Climate risk and the insurance industry

 

Many studies consider weather and climate risks to be synonymous. We use the NASA (2005) definitions of climate and weather. The main difference between the two definitions is time. Weather is atmospheric conditions over a short period of time, while climate covers a long period of time. Climate change is related to changes in average daily weather.

The potential causal impacts of new climate patterns on damages from catastrophe risks must be better estimated by the insurance industry and public authorities. These potential impacts may have been underestimated in risk management for many years. Here are some worldwide statistics obtained from the Munich Re reports of 2014, 2019, and 2021. (See also the Sigma (Swiss Re) reports (2009, 2015, 2022) and the AON reports).

 

• Climate change was rated number one among the top-ten risks facing the insurance sector (Ernst and Young, 2008).

​

• 88% of all natural catastrophic events worldwide were weather-related between 1980 and 2014 (83% in 2019); and 40% of the overall losses from 1980 to 2014 occurred in Asia (43% in 2019).

​

• 64% of insured losses were incurred in North America (incl. Central America and the Caribbean) during this period (35% in 2019), which represents about 30% of overall losses in this region, which is similar to the proportion of overall losses in the rest of the world. Insurance penetration is low, even in developed countries.

​

• Natural disasters accounted for $280 billion in economic losses around the world in 2021 ($120 billion insured). The record year was 2011, with losses of $355 billion. About 10,000 deaths were attributed to natural disasters in 2021. In the US, $145 billion in losses were observed in 2021, with $85 billion insured. In 2023, more than $100 billion US insured losses were observed from natural catastrophes. This amount was observed each year over the four last years and during six years over the last seven years (Gallagher Re, 2024).

​

The average economic losses related to natural catastrophes over the last 10 years are $187 billion ($340 billion in 2017 only). The year 2019 was below the last 10-year average, with a total loss of $150 billion, while, in 2020, the total loss was $210 billion (Munich Re). However, event frequency has increased. In 2019, there were 33 events of over $1 billion in total losses each. Nine events cost the insurance industry over $1 billion that year, and all of them were weather risk events (cyclones, storms with flooding, and tornadoes). Moreover, in April 2020, severe weather events in the U.S. cost insurers billions of dollars, with 14 tornadoes occurring that month— the fifth-highest monthly number on record since 1950, according to the Aon Global Catastrophe Recap (2021).
 

In 2021, 22 weather disasters of $1 billion or more were observed in the United States, for a total of $145 billion in damages. Since 1980, 310 events of $1 billion or more have accounted for $2.5 trillion, with an average of $148 billion per year over 2016–2021 (www.climate.gov/disasters2020).

Modeling firm AIR Worldwide now estimates that the losses to insured industry from Hurricane Ida in 2021 will be between $20 billion and $30 billion (possibly 35 billion with 67 billion economic losses, according to Munich Re). The estimate includes wind and storm surge losses of $17 to $25 billion, and private-market insured losses from inland flooding of $2.5 billion to $5 billion. These estimates include insured physical damage to residential and commercial property and autos, but do not include National Flood Insurance Program losses. Most insured losses are in the homeowner and commercial property lines of business in Louisiana and in the Northeast, including New York and New Jersey. With an estimated $30 billion in insurance losses, Ida is in the range of Hurricanes Andrew, Maria, Irma, and Harvey. State officials have reported more than 80 deaths due to Ida. Irma caused economic losses of 67 billion in 2007, with 30 billion to the insurance industry, according to Swiss Re.
 

The escalating frequency and severity of extreme weather-related events highlights a dangerous link between insurance risk and climate change, even though less than 40% of the total losses are covered. According to a Price Waterhouse Coopers survey conducted in 2017, natural catastrophes are now the second-highest risk that insurance companies face, while global warming is ranked fourth. A more recent survey by Deloitte (2020) found that most U.S. state insurance regulators expect all types of climate change risks to insurance companies to increase over the medium to long term. More than half the state regulators surveyed also indicated that climate change is likely to have a high impact on coverage availability and underwriting assumptions. U.S. state regulators and lawmakers are concerned about the insurance industry’s response to climate change. Two traditional mechanisms are usually used to reduce financial fragility: insurers can increase premiums in the states or counties most affected, or increase reinsurance coverage (Grenier, 2019). However, these two alternatives may not be sufficient to ensure the long-run stability of the industry.
 

We can summarize the major issues related to climate risks as follows (Dionne, 2015):
 

• For many years the population has concentrated in high-risk areas. This increases insurers’ exposure to major catastrophes related to natural hazards (low frequency and high severity) (Grislain-Letrémy and Villeneuve, 2019; Goussebaïle, 2016).
   

• The demand for insurance coverage for weather risk among individuals is low (Arrow, 1982; Dixon et al., 2006; Robinson and Botzen, 2022) because the potential insured underestimate the risk and are biased in estimating their potential net loss due to anticipated government intervention. For example, although flood insurance has been subsidized by the U.S. federal government since 1968, demand remains low (Kousky, 2018; Landry and Jahan-Parvar, 2011; Wagner, 2022).
   

• On the supply side, a survey funded by the National Association of Insurance Commissioners (NAIC) mentions that insurers reported increased engagement in climate-related activities over recent years, while they were not really prepared to cover weather risk in 2014 (NAIC, 2020). See also the study of Gatzert and Reichel (2022).
   

• Natural hazard losses fluctuate radically. This is a long-run issue. Insurers cannot restrict themselves to recent loss history to calculate premiums and capital. They must compute, for example, the estimated maximum loss (EML) or the expected shortfall (or CVaR), obtained from data over many years, and perform appropriate dynamic stress testing.
   

• Prevention is a long-run investment activity, yet insurance coverage is annual. This creates a problem of the insurance industry having a long-run commitment to potential investors, leading to underinvestment in prevention.
   

• Insurers can spread their liabilities through reinsurance. In principle, the effects of catastrophes can be diversified through the worldwide reinsurance market. Historically, the capacity available to reinsurers was limited, but it has increased significantly since Hurricane Andrew (Cummins and Weiss, 2000, 2004). Even though insurers and investors around the world are now more convinced that lack of action to combat climate change is becoming costly in the long run, no real structural changes have been made. The current actions intended to reduce the social costs of climate risk may not be the most efficient. In fact, some reinsurers have limited their exposure to such losses, and rating agencies seem to encourage such a move to maintain the current ratings of (re)insurance companies. Some reinsurers are more positive but argue that this new environment is very complex and that the reinsurance industry is learning how to improve its participation in these new environmental and economic realities (Kessler, 2015; Drexler and Rosen, 2022).
   

• Insurance-linked securities (ILS) are becoming important in the reinsurance market for catastrophe losses related to climate risk and earthquakes (Lakdawalla and Zanjani, 2012; Götze and Gürtler, 2022; Carayannopoulos et al., 2022). They are not very prevalent in the insurance market. ILSs can lower the cost of risk transfer in harsh (re)insurance market conditions. They help maintain (re)insurance capacity and offer multi-year protection. They limit credit risk by collateralizing losses. For investors, ILSs are noncorrelated with other market, liquidity, and credit risks, so they represent an important diversification asset. Moreover, the capitalization of financial markets is much higher than that of (re)insurance markets. ILS penetration can reduce the price of insurance in the long run and increase the demand for insurance. However, the participation of financial markets in weather risk after a major disaster is a long-run commitment issue: will they stay in such a risky market after suffering a very big loss?
   

• Securitization and market consolidation are other market mechanisms that can improve market capacity (Cummins and Weiss, 2009; Cummins and Trainar, 2009; Boubakri et al., 2008; Berger et al., 2000; Akhigbe and Madura, 2001; Cummins et al., 1999b; Cummins and Xie, 2006; Weiss and Chung, 2004; Weston et al., 2004).

​

Although estimates vary, it seems clear that a substantial gap exists between the existing reinsurance coverage and a catastrophic loss exceeding the $15–20 billion range. For example, Swiss Re (1998) estimated that reinsurers would pay 39% of a once-in-a-century catastrophe loss in the U.S., such as a $56 billion hurricane or a $65 billion earthquake in California. The Swiss Re study estimated there was a worldwide total of $53 billion in catastrophe excess-of-loss reinsurance in place in 1997. Cummins and Weiss (2000) showed that the reinsurance industry could have funded $60 billion of a $100 billion above-expected loss.

According to 2014 data, the total reinsurance capital is about $575 billion ($660 billion, 2021), including $62 billion in ILS capacity other than traditional reinsurance. Alternative capacity (ILS) includes collateral reinsurance, sidecar, industry loss warranty (ILW), and CAT bonds. As complements to reinsurance, they represented about 10% of the global catastrophe reinsurance capital in 2014 (250-year occurrence). We may think there is sufficient capacity because annual average long-run catastrophe losses are around $150 billion, but there have been significant recent exceptions: in 2011 ($375 billion), 2017 ($340 billion), and 2021 ($343 billion) (AON, 2022).

​

• Academic research on catastrophic risk and the insurance market

​

The early academic contributions agree that natural catastrophes affect the insurance market and that this effect was increasing over time because of population migration to coastal areas and the increased valuation of properties in these high-risk areas. Shelor et al. (1992) and Lamb (1995) obtain contradictory results, however, on what effect natural disasters have on the insurance industry’s profitability. Berz (1997) was one of the first to hypothesize the impact of the greenhouse effect on the insurance industry, concluding that the future of the insurance industry could be jeopardized if insurers do not adapt to the new climate conditions. He did not document the effect with data, however. Cummins et al. (2002) show that unanticipated natural events may create liquidity problems for insurance companies in the short run, and solvency problems in the long run.

In their theoretical analysis, Cummins et al. (2002) propose a sufficient condition for capacity maximization: all insurers must hold a net of reinsurance underwriting portfolio that is perfectly correlated with aggregate industry losses. Estimating capacity using insurers’ financial statement data for 1983 to 1997, they find that the industry could adequately fund a $100 billion insured loss event, whereas U.S. insurers’ equity capital was approximately equal to $370 billion. To provide an idea of the potential losses at that time, Hurricane Andrew (1992) represented a loss of $19 billion, while the Northridge earthquake (1994) cost more than $13 billion. Moreover, scenarios constructed in 1997 by catastrophe modeling firms suggest the feasibility of a $76 billion hurricane in Florida, a $21 billion hurricane in the Northeast, a $72 billion California earthquake, and a $101 billion New Madrid earthquake.

Cummins et al. (2002) also show that the industry would be able to pay very high percentages of industry losses. For example, for a $20 billion catastrophe, they estimate that the industry could have paid at least 98.6% of the insured loss in 1997. The estimated percentages paid for larger losses declines, however. For example, according to their parameter estimates, the industry would have been able to pay, in 1997, about 96.4% of a $100 billion loss based on the group sample, and 92.8% based on the company sample. For a $200 billion loss, the industry could have paid 84.0% based on the group sample, and 78.6% based on the company sample.

Dionne and Desjardins (2022) updated the study of Cummins et al. (2002) with new data available up to the end of 2020. They verified how the insurance market’s capacity has evolved over recent years. They show that the U.S. insurance industry’s capacity to pay catastrophe losses is higher in 2020 than it was in 1997. With their data, insurers could pay 98% of a $200 billion loss in 2020, compared to 81% in 1997.

Moreover, such events may cause numerous insolvencies and severely destabilize insurance markets. For instance, a $100 billion catastrophe is projected to cause 30 insolvencies for the group sample and 136 insolvencies for the company sample. The number of insolvencies at 1991 capitalization levels would have been 108 groups and 216 companies. This means that many insurers were not ready for such potential catastrophes and may have become good targets for acquisition. Their data are taken from the regulatory annual statements filed by insurers with the NAIC.

They are able to estimate insurers’ responses for different scenarios, such as a Category-5 hurricane hitting Miami or a magnitude-8.2 earthquake in San Francisco. Their measure of capacity is based on how much equity or surplus is available, and how effectively the riskiness of insurance losses is spread though the insurance market. The traditional instrument for spreading risk among insurers is reinsurance. By buying and selling options on their portfolios with each other or with specialized reinsurers, insurers can change the risk characteristics of their portfolios.

However, there is a very large number of potential catastrophe scenarios, and the data requirements to conduct such an analysis for the entire insurance industry are enormous. Moreover, while such scenarios are valuable for planning at the firm level, they do not provide enough detail to assess the risk-spreading efficiency of the total insurance market. Rather, they seek a more general response function. Cummins et al. (2002) estimate the distributional characteristics of catastrophic losses and allocate such losses to individual insurers, using correlations between losses and financial data. The result is an option-like function that defines the estimated deliverable insurance payments conditional on any given size of aggregate catastrophic loss and that projects the number of insurer insolvencies that would result.

When capital and surplus levels are high, most insurers plan to use capital to make deals. According to a recent survey by KPMG (2018), about three-quarters of insurers expect to conduct an acquisition, and two-thirds plan to seek partnership opportunities over the next three years. Eighty-one percent say they will conclude up to three acquisitions or partnerships in the same period. As a top priority, 37% hope to transform their business model, 24% want to transform their operating model, and 10% are looking to acquire new innovative capabilities and emerging technologies through their acquisitions. The key goal is to obtain a deal that generates a contribution over the next 10 to 15 years.

A.M. Best manages a database of more than 1,000 property- and casualty-insurance companies that have failed in the United States since 1969 (Kelly, 2015). The most common reasons for insolvency are deficient loss reserves, inadequate pricing, and rapid growth. Natural disasters are the seventh-most common reason, accounting for 7% of insolvencies. The Financial Services Authority (FSA) in the United Kingdom has assessed 270 property and casualty insurance companies that failed in the European Union since 1969. Many factors are identified as primary or contributing factors, with natural hazards found to have made a small contribution. Yet, in both these studies, the data cover a very long period, and it is not clear that they are representative of the last 20 years.

Regarding other pertinent contributions, Anderson and Gardiner (2008) provide a guideline to help insurance companies manage climate risk. Availability and affordability are the major problems. Insurers alone cannot effectively reduce the social cost of climate risk. More coordination with governments is necessary for prevention. Another failure is the lack of a link between sustainability and disaster resilience. Insurers must be more active in unifying green and disaster-resilience efforts in sectors such as construction, agriculture, and land use (see also Hallegate, 2014).

Mills (2009) analyzes different mechanisms that aim to improve the insurance industry’s capacity to cover insurable losses: new coverage products; a better understanding of climate change; and the financing of activities intended to reduce climate risk, including government participation when necessary. Gollier (2005) discusses in detail the necessary role played by government to reduce the fragility of the insurance industry when extreme events occur. He argues that the government should act as a reinsurer to reduce the number of bankruptcies. The government should be the reinsurer of last resort, as in the Terrorism Risk Insurance Act (TRIA) or the Price-Anderson Act. Others favor stronger private risk-management activities for natural disasters (see Mills, 2009; Michel-Kerjan, 2012, 2015; Kunreuther, 2018; Aerts et al., 2014; Collier et al., 2021; and Klein and Wang, 2009). Jametti and Ungern-Sternberg (2010) do not consider the observed risk selection between the private and public sectors as optimal in cases where the private sector keeps acceptable or lower losses, and the public sector is limited to extreme losses. Louass and Picard (2021) propose a new characterization of optimal insurance coverage for low-probability catastrophic risks. They derive determinants of insurability and socially optimal risk sharing for events that have a low probability and high severity and that affect many individuals.

Born and Viscusi (2006) take a different approach to analyzing the effect of natural catastrophes on the insurance industry. Using data from the Swiss Re Sigma Reports for the 1984–2004 period, they show that small insurers are more likely to be affected, because they are less diversified. Finally, Born and Klimaszewski-Blettner (2013) affirm that some insurers tend to reduce their activities when they are subject to severe regulations or when they receive unanticipated large claims. Such reduction-of-activities behavior is less frequent for large insurers that are better diversified.
​