DRAFT Threats to Insurability?1 Carolyn Kousky Resources for the Future [email protected]
1. Introduction Estimates of the average annual costs of natural disasters between 2000 and 2012 range between $94 billion and $130 billion (Kousky 2013a). These losses, in inflation adjusted terms, have been growing over time (e.g., Kunreuther and Michel-Kerjan 2007). Much of the increase is due to growth in exposure: more people and capital—and more valuable capital—located in risky locations (e.g., Pielke et al. 2003; Miller et al. 2008; Barthel and Neumayer 2012; Aon Benfield 2013). Climate signals are also emerging in some loss data (e.g., Sander et al. 2013) and losses are projected to increase as the climate warms (e.g., Ranson et al. 2014). The largest share of disaster costs—at roughly 85%— comes from atmospheric and hydrologic events (Cutter and Emrich 2005; Gall et al. 2011); they are the events most likely to be impacted by climate change. Insurance is an important risk management tool for dealing with the cost of disaster events. It can protect people against risks that have a small probability, but which could be financially devastating. When households and businesses are insured, they are more likely to rebuild (Turnham et al. 2011), which can limit negative economic multiplier effects in a community. Also, by limiting the exposure of firms or households to certain risks, insurance can allow beneficial economic activity to occur that might otherwise not be undertaken. Natural disasters, however, are challenging to insure and breakdowns in disaster insurance markets are not uncommon. Low take-up rates for disaster insurance are observed, from flooding to earthquakes (e.g., Palm 1995; Dixon et al. 2006). In some locations and for some perils, private insurance companies have ceased to write policies or otherwise limited their underwriting (e.g., Stroud 2012). Governments have intervened in these markets in a wide variety of forms around the world, from writing disaster policies directly, acting as a reinsurer, backstopping disaster losses, and providing disaster assistance, among other types of interventions (e.g., Kousky 2011; Swiss Re 2011). Insurance may become even more challenging for natural disasters, as a result of many converging trends. Climate change altering weather-related extreme events is one, along with new technologies being deployed at a rapid rate, increasing integration of supply chains, vast mobility of the population, and capital continuing to concentrate in risky areas. As the world grows hotter, more integrated, and more crowded, we may need to rethink our risk management approaches and invest in a more thoughtful and targeted way in risk mitigation. This will require responsibilities and complementary efforts by both the public and private sectors.
An initial version of this chapter was written in advance of an event at Resources for the Future that the author introduced and moderated titled “Limits to Securitization.” The panel discussion is available for online viewing: http://www.rff.org/events/event/2014-06/limits-securitization-future-insurance. I would like to thank the panelists in that event for sharing their insights: Roger Cooke, Andrew Castaldi, Bob Kopp, Bob Litterman, and Peter Nakada.
DRAFT The next section of this chapter discusses the conditions that make a risk more or less easy to insure. The third section outlines several emerging challenges to the insurability of disaster risks: (1) greater dependence, particularly in the tails; (2) fat tails, getting fatter; and (3) larger systemic risks. The fourth section offers suggestions on improving the insurability of disasters. 2. The Insurability Spectrum Insurance is a form of risk transfer. A risk-averse entity is willing to pay more than the expected loss to transfer a risk to another entity better able to spread the risk. For risk transfer to be possible, several conditions must be met; these are often presented as the ideal conditions for insurability. Some are straightforward. The loss distribution must be known to a reasonable degree to make underwriting possible. The loss must be uncertain and, to some extent, random. It is not possible to insure an event known to occur with certainty or over which the insured has full control. In addition, losses must be, to a large extent, measurable and verifiable after a disaster event. The risk should not be subject to high levels of adverse selection or moral hazard. Adverse selection occurs when the insured knows more about their risk than the insurance company and only high-risk households insure. Moral hazard occurs when an insured fails to take actions to reduce their risk since damages will be compensated by the insurance company. Both of these drive up the costs of insurance—in the extreme, to levels that make underwriting impossible. In addition, ideal insurance conditions include that the loss distribution be sufficiently thin-tailed and claims be independent. What do these two conditions mean? The first is that the probability of a loss becomes negligible at extreme values. For instance, height is thin-tailed: at greater heights the probability of occurrence becomes negligible. That is, we do not see nine-foot people. Independent risks imply that when one person suffers a loss, others do not necessarily suffer a loss, as well. For example, when an individual gets in a car accident, it does not mean that their neighbors will also be in accidents. These conditions are what make risk pooling beneficial. When risks are thin-tailed and independent, the average claim of all the policies approaches the expected value. Finally, risks will only be insured if there is a price that is profitable for the insurance company and for which the insured is willing and able to pay. The market must clear. Natural disasters can violate all of the last three criteria. Damages from many disasters are decidedly fat-tailed (e.g., Holmes et al. 2008; Wildasin 2008). This means the most severe event observed is many multiples of the second most severe event. With fat-tailed losses, the probability of damage falls slowly relative to the severity. Damages from disasters are also spatially correlated, such that when one house suffers damage, it is likely all their neighbors did, too. This combination of fat tails and spatial correlation will make losses much more volatile from year to year and in some years, very severe losses will occur. For catastrophe risks, therefore, insurance firms must solve an intertemporal smoothing problem of trying to match regular premium payments, insufficient in any given year to cover a large loss, with the need for enormous sums of capital in the catastrophe years (Jaffee and Russell 1997). Given this, firms managed to keep the probability of insolvency below a certain level (value-atrisk requirements) will be required to charge large premiums to cover catastrophic risks. This can lead to the violation of the third criterion: this high price may not be one at which there is any demand for insurance (Kousky and Cooke 2012).
DRAFT Insurability of risks, though, is a spectrum. Some are easy to insure, like auto accidents, while some present so many challenges, they will not be insured at all by the private sector, such as the risk of a nuclear terrorist attack. There are many strategies that insurance companies can employ to make a risk more insurable. Insurance companies insure themselves through reinsurance, which is able to achieve better diversification across geographic regions and perils. Firms may limit their exposure through selective underwriting; for example, limiting the number of certain policies in certain locations. Insurance contracts may be modified, such as through higher deductibles (as is the case with hurricanes in much of the US) or caps on coverage. Insurability is also a dynamic concept, changing over time (Swiss Re 2005). Risks change, economic conditions change, new technologies emerge, and new information becomes available, all of which could impact the insurance market (Kousky 2013b). 3. Emerging Threats to Insurability This section discusses three emerging threats to insurability: (1) greater dependence between claims, particularly in the tails; (2) loss distributions with fattening tails; and (3) the emergence or strengthening of systemic risks. 3.1. Dependence in claims Natural disasters can generate many types of dependence across insurance claims. First, as mentioned, is simple spatial correlation in losses. When a disaster event occurs, it impacts an entire community or region. More troubling, dependence between risks can concentrate in the tails, referred to as tail dependence. Tail dependence can be defined as the probability that one variable takes on an extreme value conditional on another variable taking on an extreme value. Some insurance lines have been found to be tail dependent. For example, automobile, health, property, and life insurance claims may usually be fairly independent, but in a severe disaster event, they all suffering losses simultaneously, as observed after severe storms (RMS 2005; Lescourret and Robert 2006). It is possible that climate change could introduce or strengthen dependencies between risks, including tail dependence (Kousky and Cooke 2009). This could be a simple function of more frequent or more severe disaster events that are more likely to have multiple impacts. For example, extreme heat is projected to increase as the climate warms and we know that heat waves can have a range of negative outcomes. The 2003 heat wave in France, for example, led to uninsured crop losses, fires, loss of nuclear power, higher electricity prices, rockfalls, and excess mortality from the heat and higher groundlevel ozone (De Bono et al. 2004; Schär and Jendritzky 2004; Stedman 2004). Concentration of exposure in hazardous areas can also increase dependence in claims arising from the occurrence of natural disasters. For example, the severe 2011 floods in Thailand highlighted correlations in contingent business interruption losses due to a concentration of manufacturing in the floodplains of a single country. Many companies were not even aware of this risk in their supply chain and (re)insurance companies saw claims come in from around the world as the business interruption cascaded, particularly through the technology and automotive manufacturing sectors. The event also led to claims being filed for much longer than usual as all the interconnections in supply chains were slowly revealed (Holbrook 2013). Failure to recognize dependencies can lead to a failure of risk management. While there is a saying in finance that in a time of crisis, all things are correlated, those possible tail correlations need to 3
DRAFT be adequately assessed ex-ante if they are to be managed. Yet assessing the dependencies between risks can be quite challenging, particularly for tail dependence where many observations of extremes are required to statistically identify the relationship. It has been argued that a failure to appreciate the increased complexities in the financial markets was partially responsible for the economic crisis (Goldin and Vogel 2010). Recently, structured expert judgment has been brought to bear on improving modeling of dependence (Cooke and Goossens 2000; Morales et al. 2007). Risk Management Solutions, a catastrophe modeling firm, has also noted there may be triggers, which they refer to as phase transitions, where a system moves from roughly independent damages to highly correlated damages. One example they give is a forced evacuation, after which properties deteriorate, there is a lack of personnel for response and operation of critical facilities, and rebuilding costs rise (RMS 2005). Clearly, more work is needed in identifying and quantifying dependence, when it emerges, and how it may be changing in response to climate, demographic, or technological trends. 3.2. Fatter tails As already noted, many disasters have been found to be fat tailed. Climate models predict that the frequency, magnitude, location, and/or duration of many extreme events may be changing (IPCC 2012). These climate changes may be shifting the distribution of disaster losses to the right, leading to more extreme events (U.S. Climate Change Science Program 2008). There is also concern that the distribution of disaster impacts may not just be shifting, but the tails may also be getting fatter. Initial evidence of this has been found in flood insurance claims, although the causes are unknown and likely to involve concentrations of development in high risk areas (Cooke et al. 2014). As another example, hurricane strength may be increasing in response to warmer sea temperatures (Knutson et al. 2010). Other global changes may be fattening the tails of disaster loss distributions, as well. As one example, the dramatic increase in the mobility of the population, as well as commodities, worldwide has implications for the spread of disease (Knobler et al. 2006). Mobility changes are coupled with climate changes that are allowing diseases to enter regions where they were previously not found. These coupled trends mean that previously localized disease outbreaks can quickly become global pandemics. As another example, leaner and more integrated supply chains potentially produce economic gains, but also can increase society’s vulnerability by fattening the tail of loss distributions. Food supply chains have moved toward much smaller inventories, concentrated among a smaller number of suppliers, bringing in food from around the globe instead of locally (Mahanta 2013). This means that there is a greater chance for disruption in the chain and when a disruption does occur, it has a bigger impact as there is very little cushion. Essentially, there has been a move toward elimination of redundancy in supply chains in the name of efficiency. While potentially improving earnings in stable periods, this can lead to an escalation of damages when a system is stressed or faced with a threat. A similar push to eliminate “idle” capital and to seek leverage from capital, even if reserved for risk management, contributed to the financial crisis of 2008 (Goldin and Vogel 2010). If systems have a “buffer,” it can lower damages when an extreme event occurs. The challenge is that such buffers appear wasteful in non-disaster times. Both tail dependence and fatter tails can increase the “probable maximum loss” (PML) of an insurance company. The PML is an estimate of the worst-case losses that an insurance company could face, either for a policy or for an entire portfolio. Increases in the PML will require increases in capital or 4
DRAFT will increase the solvency risk insurers face. Presumably such increases would result in higher insurance premiums, which, as already stated, could lead to policies costing more to write than insureds are willing or able to pay. 3.3 Larger systemic risks Systemic risks—a term generally used to refer to a system-wide risk in which an event affects all entities simultaneously—are, by definition, non-diversifiable and thus not insurable. For example, a company cannot purchase insurance against a global recession. This, in and of itself, is nothing new. What is concerning is the possibility that we are introducing more systemic risks across sectors. There have been efforts to tease lessons from managing systemic risks across fields (e.g., National Research Council 2007), which could inform such changes. Different definitions of systemic risk are found in the literature. In general, however, there seem to be two types of systemic risks: (1) a very large event that by its magnitude effects all or most entities in a system, and (2) an event of any size that sets in motion a cascade or chain reaction of negative consequences that ultimately impacts most or all of a system (e.g., Kaufman and Scott 2003). In a sense, these are the extreme of the two challenges just discussed—fat tails and dependence—in the former, if an event is fat tailed enough, then an occurrence could be large enough to impact the entire system and in the latter, if risks are highly dependent, negative consequences could cascade and again the effect could be systemic. Both types of systemic risk may be evolving with recent global trends. Regarding the first type of systemic risk, scientists are concerned, for example, that climate change could lead to tipping points in the earth system, or other catastrophic impacts, which would induce global, that is systemic, losses (e.g., Schneider 2004; AAS and RFF 2015). Consider collapse of the thermohaline circulation in the Atlantic, or melting of a section of the West Antarctic Ice Sheet, which would cause at least four feet of global sea-level rise (Joughin et al. 2014; Rignot et al. 2014). This sealevel rise is not insurable as it is inevitable and global (albeit with some local variation). Regarding the second type of systemic risk, there is also the possibility that climate change could lead to cascading consequences which turn a localized event into a global one (e.g., Kousky et al. 2009). For example, a group of retired U.S. admirals and generals found that climate change could be a “threat multiplier” and lead to instability in volatile regions (CNA Corporation 2007). Of course, what constitutes a systemic risk depends on the system under study. For a small insurance company writing policies in only one region, a single hurricane could be a systemic risk. Consider the small, state-based, so-called “take-out” companies in Florida that are paid to remove policies from the state wind pool. There is concern these firms, with such concentrated exposure, could not withstand a large hurricane; indeed, many have failed even absent a hurricane (Olorunnipa 2013). For a global reinsurance company, however, it would take a global event to reach systemic levels. Global diversification is thus an important tool for managing smaller scale impacts, but some threats, as discussed, stress even the global reinsurance sector and it is those that could prevent the use of insurance as a risk management tool. 4. Can We Make the Uninsurable Insurable? 4.1 Targeted hazard mitigation 5
DRAFT When the risk of extremes is lower, insurance is more feasible and is less costly. A natural first response to challenges of insurability, then, is to examine the possibility for investments in hazard mitigation that could thin the tail of the loss distribution. This possibility drove the creation of the Manufacturers Mutual Fire Insurance Company in 1835. The company only covered properties that had engaged heavily in risk reduction measures and offered lower rates to those properties as a result (Swiss Re 2005). This ultimately became the firm FM Global which still employs this strategy. Can we also identify interventions that could prevent cascading consequences or contagion effects? For example, are there “circuit breaker” climate adaptation or hazard mitigation strategies (Kousky et al. 2009) that could be employed to manage systemic threats? Similarly, could we identify investments that decrease problematic correlations? For example, take the 1906 earthquake in San Francisco. This event showed the tail dependence between earthquakes and fires (Steinberg 2000). Breaking this dependence involves the design and installation of gas and water pipes that can withstand extreme earthquakes. Another example is President Obama’s recent suggestion to designate utility workers as first responders. This is an official designation from FEMA that would allow access to disaster sites. Power outages can prevent important relief and repair work or even exacerbate damages and lead to cascading losses. Designating lineman as first responders can help bring the power back faster and reduce damages. A similar expansion could be made for those who help restore internet connectivity or other communication networks. Risk reduction often must be a partnership between the public and private sectors. Insurance could provide some incentive for individuals and businesses to invest in hazard mitigation. If premiums are lowered when risks are reduced, actions that cost less than this difference will be financially attractive for the insured (assuming credit availability). As noted by Woo (1999), however, insurance may not be the best driver of hazard mitigation since in soft market conditions, when prices are low, it can discourage risk reduction, and when prices are high, many forgo disaster insurance entirely. There is also a large literature suggesting individuals may not evaluate risks as an expert would and thus may fail to adopt cost-effective mitigation measures. Government, in its capacity to regulate land use, enact building regulations, and fund large structural and nature infrastructure investments may need to promote risk reduction, thus helping making the residual risk insurable. 4.2. Models beyond traditional insurance In recent decades, several innovative risk transfer mechanisms have emerged that could complement traditional insurance in providing cover for difficult-to-insure risks and/or access the capital in financial markets. After Hurricane Sandy, the New York City Metropolitan Transit Authority faced high insurance prices and opted instead for a $200 million catastrophe bond. The bond relies on a parametric trigger—the height of storm surge at tide gauges. More entities may opt for this type of disaster financing, as it can provide more flexible and possibly cheaper access to capital. The risk of parametric based products, however, is that the actual loss will not perfectly match the payout: it may be higher, but it could also be lower. That said, this creates in incentive for hazard mitigation. As has been recommended for covering flood events in the United States (Michel-Kerjan and Kunreuther 2011), taking a layered approach to disaster risks could prove most effective. For the first layer of losses, the individual or business will retain the risk and self-insure. To lower the costs of insurance, this deductible could be as high as the entity can comfortably manage. For firms, this could 6
DRAFT involve creating a captive insurance company to be able to set aside capital to cover disaster events, and for individuals could involve the use of disaster savings accounts. The next layer of losses could be covered by standard insurance or through pools of similar entities, such as municipal pools in the United States. After this, losses can be reinsured and/or placed in the financial markets using insurance linked securities. Finally, for the most severe events, the government can act as a backstop. This could also involve a layer of government reinsurance, such as is done in the U.S. for terrorism, particularly for higher levels of loss where tail dependence is problematic. However structured, a layered approach would divide up losses among entities best able to handle the given level of risk. 5. Conclusion As a changing climate and increasingly interconnected economy combine to create the potential for greater extreme events, traditional insurance markets could potentially become stressed for particular perils or in particular regions. Governments around the world have already taken a variety of approaches to helping disaster insurance markets, from offering coverage, to providing reinsurance, to simply helping regulate building in a way that minimizes losses. Insurability of risks is not a yes-no proposition, however, but a continuum and there are many actions that can be taken to make risks more insurable by all sectors. Partnerships between the public and private sector to continue to manage changing risks will continue to be critical to promoting resilience at a range of scales. References RFF and AAAS (2015). The Economic and Financial Risks of a Changing Climate: Insights from Leading Experts. Workshop Report. Washington, DC: Resources for the Future, November 12. Aon Benfield (2013). Annual Global Climate and Catastrophe Report Impact Forecasting — 2012. Chicago, Illinois, Impact Forecasting and Aon Benfield, Barthel, F. and E. Neumayer (2012). "A trend analysis of normalized insured damage from natural disasters." Climatic Change 113: 215-237. CNA Corporation (2007). National Security and the Threat of Climate Change. Alexandria, Virginia, The CNA Corporation, Cooke, R. and L. J. H. Goossens (2000). Nuclear science and technology. Procedures guide for structured expert judgment. EUR 18820EN. Luxembourg, European Commission, Cooke, R., C. Kousky and E. Michel-Kerjan (2014). Flood Insurance Claims: A Fat Tail Getting Fatter. Common Resources. Washington, DC, Resources for the Future, April 24. Cutter, S. L. and C. Emrich (2005). "Are Natural Hazards and Disaster Losses in the U.S. Increasing?" EOS, Transactions, American Geophysical Union 86(41): 381, 388-389. De Bono, A., G. Giuliani, S. Kluser and P. Peduzzi (2004). Environmental Alert Bulletin: Impacts of Summer 2003 Heat Wave in Europe. Châtelaine, Switzerland, United Nations Environment Programme, Global Resource Information Database (GRID) - Europe Dixon, L., N. Clancy, S. A. Seabury and A. Overton (2006). The National Flood Insurance Program’s Market Penetration Rate: Estimates and Policy Implications. Santa Monica, California, RAND Corporation, February. Gall, M., K. A. Borden, C. T. Emrich and S. L. Cutter (2011). "The Unsustainable Trend of Natural Hazard Losses in the United States." Sustainability 3: 2157-2181. 7
DRAFT Goldin, I. and T. Vogel (2010). "Global Governance and Systemic Risk in the 21st Century: Lessons from the Financial Crisis." Global Policy 1(1): 4-15. Holbrook, E. (2013). A World of Supply Chain Risks. Risk Management, April 9. Holmes, T. P., R. J. Huggett and A. L. Westerling (2008). Statistical Analysis of Large Wildfires. The Economics of Forest Disturbances: Wildfires, Storms, and Invasive Species. T. P. Holmes, J. P. Prestemon and K. L. Abt, Springer Science: 59-77. IPCC (2012). Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. C. B. Field, V. Barros, T. F. Stockeret al. Cambridge, UK, Cambridge University Press, Jaffee, D. M. and T. Russell (1997). "Catastrophe Insurance, Capital Markets, and Uninsurable Risks." The Journal of Risk and Insurance 64(2): 205-230. Joughin, I., B. E. Smith and B. Medley (2014). "Marine Ice Sheet Collapse Potentially Under Way for the Thwaites Glacier Basin, West Antarctica." Science 344(6185): 735-738. Kaufman, G. G. and K. E. Scott (2003). "What is Systemic Risk, and Do Bank Regulators Retard or Contribute to It?" The Independent Review VII(3): 371-391. Knobler, S., A. Mahmoud, S. Lemon and L. Pray, Eds. (2006). The Impact of Globalization on Infectious Disease Emergence and Control: Exploring the Consequences and Opportunities, Workshop Summary - Forum on Microbial Threats. Washington, DC, The National Academies Press. Knutson, T. R., J. L. McBride, J. Chan, K. Emanuel, G. Holland, C. Landsea, I. Held, J. P. Kossin, A. K. Srivastava and M. Sugi (2010). "Tropical cyclones and climate change." Nature Geosci 3(3): 157163. Kousky, C. (2011). "Managing the Risk of Natural Catastrophes: The Role and Functioning of State Insurance Programs." Review of Environmental Economics and Policy 5(1): 153-171. Kousky, C. (2013a). "Informing climate adaptation: A review of the economic costs of natural disasters." Energy Economics http://dx.doi.org/10.1016/j.eneco.2013.09.029. Kousky, C. (2013b). Revised Risk Assessments and the Insurance Industry. RFF Discussion Paper 13-35. Washington, DC, Resources for the Future, Kousky, C. and R. Cooke (2009). Climate Change and Risk Management: Challenges for Insurance, Adaptation, and Loss Estimation. RFF Discussion Paper 09-03-REV. Washington, D.C., Resources for the Future, Kousky, C. and R. Cooke (2012). "Explaining the Failure to Insure Catastrophic Risks." Geneva Papers on Risk and Insurance - Issues and Practice 37(2): 206-227. Kousky, C., O. Rostapshova, M. Toman and R. Zeckhauser (2009). Responding to the Threats of Climate Change Mega-Catastrophes. RFF Discussion Paper 09-45. Washington, DC, Resources for the Future, Kunreuther, H. C. and E. Michel-Kerjan (2007). "Climate Change, Insurability of Large-Scale Risks and the Emerging Liability Challenge." Penn Law Review 155(6): 1795-1842. Lescourret, L. and C. Y. Robert (2006). "Extreme Dependence of Multivariate Catastrophic Losses." Scandinavian Actuarial Journal 4: 203-225. Mahanta, S. (2013). New York's Looming Food Disaster. The Atlantic Cities. October 21,
DRAFT Michel-Kerjan, E. and H. Kunreuther (2011). "Redesigning Flood Insurance." Science 333(22 July): 408409. Miller, S., R. Muir-Wood and A. Boissonnade (2008). An exploration of trends in normalized weatherrelated catastrophe losses. Climate Extremes and Society. H. F. Diaz and R. J. Murnane. Cambridge, UK, Cambridge University Press: 225-247. Morales, O., D. Kurowicka and A. Roelen (2007). "Eliciting Conditional and Unconditional Rank Correlations from Conditional Probabilities." Reliability Engineering and System Safety 93: 600710. National Research Council (2007). New Directions for Understanding Systemic Risk: A Report on a Conference Cosponsored by the Federal Reserve Bank of New York and the National Academy of Sciences. Washington, DC, The National Academies Press. Olorunnipa, T. (2013). "Despite No Hurricanes, Many "Takeout" Insurers Fail." Miami Herald, Miami, FL, June 3. Palm, R. (1995). "The Roepke Lecture in Economic Geography - Catastrophic Earthquake Insurance: Patterns of Adoption." Economic Geography 71(2): 119-131. Pielke, R. A., Jr., J. Rubiera, C. W. Landsea, M. L. Fernandez and R. Klein (2003). "Hurricane Vulnerability in Latin America and The Caribbean: Normalized Damage and Loss Potentials." Natural Hazards Review 4(3): 101-114. Ranson, M., C. Kousky, M. Ruth, L. Jantarasami, A. Crimmins and L. Tarquinio (2014). "Tropical and extratropical cyclone damages under climate change." Climatic Change DOI 10.1007/s10584014-1255-. Rignot, E., J. Mouginot, M. Morlighem, H. Seroussi and B. Scheuchl (2014). "Widespread, rapid grounding line retreat of Pine Island, Thwaites, Smith and Kohler glaciers, West Antarctica from 1992 to 2011." Geophysical Research Letters: 2014GL060140. RMS (2005). Hurricane Katrina: Profile of a Super Cat: Lessons and Implications for Catastrophe Risk Management. Newark, California, Risk Management Solutions, Sander, J., J. Eichner, E. Faust and M. Steuer (2013). "Rising variability in thunderstorm-related U.S. losses as a reflection of changes in large-scale thunderstorm forcing." Weather, Climate, & Society doi:10.1175/WCAS-D-12-00023.1. Schär, C. and G. Jendritzky (2004). "Hot News From Summer 2003." Nature 432(December 2): 559-560. Schneider, S. H. (2004). "Abrupt Non-Linear Climate Change, Irreversibility and Surprise." Global Environmental Change 14: 245-258. Stedman, J. R. (2004). "The Predicted Number of Air Pollution Related Deaths in the UK During the August 2003 Heatwave." Atmospheric Environment 3(8): 1087-1090 Steinberg, T. (2000). Acts of God : The Unnatural History of Natural Disaster in America. New York, Oxford University Press. Stroud, M. (2012). "As weather gets biblical, insurers go missing." Reuters, Pittsburgh, April 11. Swiss Re (2005). Innovating to Insure the Uninsurable. Sigma. Zurich, Swiss Reinsurance Company, 1 September. Swiss Re (2011). State Involvement in Insurance Markets. Sigma. Zurich, Switzerland, Swiss Reinsurance Company Ltd.,
DRAFT Turnham, J., K. Burnett, C. Martin, T. McCall, J. Randall and J. Spader (2011). Housing Recovery on the Gulf Coast, Phase II: Results of Property Owner Survey in Louisiana, Mississippi, and Texas. Washington, DC, U.S. Department of Housing and Urban Development, Office of Policy Development and Research, August. U.S. Climate Change Science Program (2008). Weather and Climate Extremes in a Changing Climate. Washington, D.C., U.S. Climate Change Science Program, June. Wildasin, D. E. (2008). "Disaster Policies: Some Implications for Public Finance in the U.S. Federation." Public Finance Review 36(4): 497-518. Woo, G. (1999). The Mathematics of Natural Catastrophes. London, Imperial College Press.