Sie sind auf Seite 1von 34

32

Insurance and Financial Risk Transfer


32.1 Introduction 32.2 Insurance and the Insurance Industry
The Law of Large Numbers Insurability Conditions Types of Insurance Reinsurance

32.3 Earthquake Insurance


Insurance Supply and Insurers Decision Processes Insurance Demand and Homeowners Decision Processes

32.4 Earthquake Insurance Risk Assessment


California Department of Insurance Probable Maximum Loss CRESTA Zonation Probabilistic Earthquake Insurance Loss Estimation Dynamic Financial Analysis

32.5 Government Earthquake Insurance Pools

Charles Scawthorn
Consulting Engineer Berkeley, CA

California and the California Earthquake Authority Japan and the Japan Earthquake Reinsurance Company Turkey and the Turkish Catastrophic Insurance Pool New Zealand and the Earthquake Commission

Howard Kunreuther
University of Pennsylvania Philadelphia, PA

32.6 Alternative Risk Transfer


Index-Based Cat Bonds Other Types of Cat Bonds Cat Bond Example ART Summary

Richard Roth, Jr.


Consulting Actuary Huntington Beach, CA

32.7 Summary Dening Terms References Further Reading

32.1 Introduction
Earthquakes cause damage. It is usually considered not economically feasible to construct buildings and structures so as to completely eliminate the potential for any future damage. In fact, a certain amount of damage is predicated in the formulation of building codes, as exemplied in the rst edition of the Recommended Lateral Force Requirements and Commentary [SEAOC, 1999]. This code proposed that structures are designed to meet the following multitiered performance capabilities: Resist minor earthquake shaking without damage Resist moderate earthquake shaking without structural damage but possibly with some damage to nonstructural features Resist major levels of earthquake shaking with both structural and nonstructural damage but without endangerment of the lives of occupants
2003 by CRC Press LLC

32-2

Earthquake Engineering Handbook

In addition, modern buildings may sustain large losses beyond those anticipated in the building code because of the uncertainty associated with earthquakes and errors in the design process. Finally, until recently, earthquake codes only addressed the basic structure, and did not seek to prevent damage to contents, or prevent disruption of function,1 which can result in signicant nancial loss. Given that the potential exists for large losses even in well-designed buildings, individuals and organizations often protect themselves against this nancial loss by purchasing insurance. In this chapter, we focus on how insurers approach the earthquake problem to set fair and reasonable rates of coverage against earthquakes. We additionally show under what circumstances a risk is insurable and how a company can develop new strategies for coping with catastrophic losses. We begin by explaining the basics of insurance and then examine specic risk assessment tools (exceedance probability curves and probable maximum loss estimates) used by insurers to measure earthquake risk, the insurers decision process in offering coverage against damage from earthquakes, and the risk bearers decision process in deciding whether or not to purchase coverage. We then discuss the typical earthquake insurance situation in high-risk regions, such as California, Japan, and Turkey, and conclude with a discussion of recent alternatives to traditional insurance.

32.2 Insurance and the Insurance Industry


Insurance is a contract binding one party to indemnify a second party against specied loss in return for premiums paid. Insurance is one of the principal mechanisms used to manage risk. Insurance enables one to pay a relatively small premium today to protect oneself against an uncertain but potentially large loss in the future. Basically, the insurance contract is the substitution of a certain small loss for the possibility of a large uncertain loss. Insurance rewards those who take protection against a potential loss and provides funds to cover the cost after a disaster occurs. In the United States, some insurance is required while other coverage is voluntary. For example, new homeowners are often compelled to purchase homeowners coverage as a requirement for a mortgage. In this way, the bank or other nancial institution protects its investment in case of a re or other large loss. Also, automobile liability insurance is required in most states as a condition for operating a car. Our focus in this chapter is on earthquake insurance to cover losses to property. Property insurance has its origins in re insurance and, in the English-speaking world, can be traced back to the Great Fire of London in 1666 [Kunreuther and Roth, 1998]. It is based on the collection of premiums from the members of the pool from which those suffering a loss are compensated. Property insurance works best when there is a signicant number of randomly occurring, independent losses that are large for any individual in the pool but small relative to the total insured value covered by the pool. Although there may be considerable variability between individual claims, the variance of aggregate losses is small if the events are independent as a result of the central limit theorem, or law of large numbers, discussed later. The insurance industry is one of the largest industries in the United States and the world. It is divided into two major sectors: life/health and property/casualty (P&C). The P&C industry is divided into two major markets: personal lines and commercial lines. Personal lines are the types of insurance (e.g., homeowners, auto), referred to as lines of business, sold to individuals to protect their personal property. Commercial lines are the types of insurance offered to organizations, such as companies, corporations, municipalities, etc., and include re, theft, business interruption (BI), loss of valuable papers, etc. Earthquake insurance, as well as re and most kinds of insurance protecting real property, is written by the property/casualty sector of the industry. In 1998, the P&C sector accounted for 41% of the total U.S. insurance industry, with gross direct premiums of $368 billion [Insurance Information Institute, 2002], i.e., the property/casualty industry accounted for about 4% of the entire U.S. gross domestic product.
1 Function is the ultimate goal of a facility. For most normal buildings, the function is the housing of people and their goods. For example, if the function of a factory is disrupted, products cannot be manufactured and prots are therefore not made. In the insurance industry, these lost prots are covered under a business interruption (BI) policy.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-3

32.2.1 The Law of Large Numbers


Ratemaking is the process of determining what premium (rate) to charge on a property. The premium that you pay is determined by an actuary at the insurance company. Ratemaking is normally based on the law of large numbers. Insurance markets can exist because of this. In simple terms, the law of large numbers states that, as a series of independent events under consideration increases, the distribution of the frequency of those events tends toward the normal distribution. Using re insurance as an example, as long as the occurrence of a re in one house is independent of the occurrence of a re in a second house, then the average (mean) loss per year of a collection or portfolio of n houses can be estimated with increasing accuracy for an increasing number of houses. The mean is the sum of the means, and the variance is the sum of the variances, which implies that the mean increases linearly with n houses but the standard deviation increases with the square root of n: L= and ( L) = where L = the expected value of loss for the n houses li = the expected loss for the ith house, equal to the value of that house times the probability of loss for the house = the standard deviation (a measure of the dispersion about the mean) n = summation over n Equation 32.2 is based on independence of loss between the risks. As n increases, it results in a lower standard deviation (i.e., a robust mean), or a statistical predictability that permits a limited amount of capital to insure (or protect) many properties simultaneously. This is illustrated in Figure 32.1, where the distribution of the probability of the loss (probability density function, pdf) for a small portfolio is shown on the left. Note that it has a relatively broad shape, a large standard deviation. On the right in Figure 32.1 is the pdf for a large portfolio, the expected loss (i.e., the most likely value of loss) is larger but the shape is tighter, a relatively smaller standard deviation. The expected loss of the large portfolio is less uncertain; it is better known, more predictable (on average), and therefore a more known risk. It is a larger risk but more certain, and therefore more manageable. For example, if an insurance company has an amount of capital equal to the value of a house, and if on average one house in a thousand is destroyed by re per year, then if one thousand homeowners each pay 0.1% of the value of their house per year to an insurance company, the company can use the collected payments (premiums) to pay the one homeowner whose house is totally destroyed by re, and still have a surplus of capital (its initial capital) in case, unusually, a second house is destroyed by re. The
probability

(l )
n i 2 0.5

(32.1)

(L l )
n i

(32.2)

Loss

FIGURE 32.1 Probability density function (pdf) of small portfolio (left) and large portfolio (right).
2003 by CRC Press LLC

32-4

Earthquake Engineering Handbook

determination of the average loss per year, or the loss or burning rate, is the province of actuaries, who base their estimates on loss statistics (when a large body of experience data is available), or on rational analysis and judgment (when insufcient data are available). This means that for a large number of claims, you can estimate the proper premium rate very accurately. It does not mean that the E[x] goes to zero or that insurance will become cheaper by the law of large numbers. Exactly the same principle is used in political polling to estimate the outcome of the next election, based on a sample poll. The sample does not have to be very large; a sample of 2000 is quite large. Similarly, an insurance company does not have to be very large to be able to predict the proper rate fairly accurately. The actuary estimates the average loss cost and then increases the number for future ination and company expenses. This produces the needed rate that is charged to the public. As the number of houses in the companys portfolio grows, the ratio of surplus required to protect against a greater than average number of losses in a year becomes smaller. In this manner, the company can accumulate more capital each year, a portion of which is retained to increase the surplus and thus permit growth of the companys business. When a large number of properties are affected simultaneously by a common cause, such as a conagration, ood, hurricane, earthquake, etc., then independence is lost, i.e., Equation 32.2 is not appropriate, and correlation between risks must be accounted for. This implies that many insureds must be indemnied simultaneously, leading to extraordinary demands on the companys surplus. In the insurance industry, such events are termed catastrophes, and may result in insolvency. To avoid this, the actuarial and underwriting elements of an insurance company need to constantly monitor the potential for catastrophic occurrence, or portfolio risk. In addition to the problems of insuring such high-variance losses, there is the additional problem of estimating the parameters of the probability distribution of losses. Due to the frequency of these events, it takes an extraordinarily long history of past disasters to estimate the mean loss with an acceptable degree of predictability. This is why hazard experts and risk assessors would like to have databases of earthquakes and other hazards over thousands of years. Given the relatively short period of accurate history of hundreds of years, the mean loss and other parameters of the loss distribution are difcult to estimate with an acceptable degree of accuracy. Thus, most estimates of mean loss are supplemented by scientic estimates of the risk from such experts as seismologists, meteorologists, and structural engineers (discussed in other chapters of this volume).

32.2.2 Insurability Conditions


What does it mean to say that a particular risk is insurable? We address this question from the vantage point of a supplier of insurance who offers coverage against a specic risk at a certain premium to a potential policyholder. The policyholder is protected against a prespecied set of losses dened in an insurance contract. Two conditions must be met before insurance providers are willing to offer coverage against an uncertain event: 1. The rst condition is the ability to calculate expected losses to the insurer. This includes identifying and estimating the probability of uncertain events occurring and the associated losses when providing different levels of coverage. 2. The second condition is the ability to set premiums for each potential customer or class of customers. This requires some knowledge of the customers risk in relation to others in the population of potential policyholders. If conditions 1 and 2 are both satised, a risk is classied as insurable. But it still may not be protable. In other words, it may be impossible to specify a rate for which there is sufcient demand and incoming revenue to cover the development, marketing, and claims costs of the insurance and yield a net positive prot. In such cases, the insurer will opt not to offer coverage against this risk.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-5

Condition 1: To satisfy the rst condition of insurability, probabilistic loss estimates must be calculated for various coverage levels. Such estimates can be based on historical data from previous events and on scientic forecasts of what is likely to occur in the future. Specically, for earthquakes, hazard risk maps can be utilized. Such maps have been drawn in the United States but they only provide rough guidelines as to the likelihood and potential risk of certain events. Loss data for earthquakes are quite uncertain and cover a relatively short time horizon. Therefore, scientists have been working to reduce the uncertainty in the risk assessment process for earthquakes. Such analyses, aided by new developments in earthquake risk modeling, make it easier to estimate the risk facing insurers when they are considering how much coverage to provide in hazard-prone areas. Condition 2: Once the risk has been identied, the insurer must determine a strategy to avoid an unacceptable level of earthquake loss while still making a prot. There are a number of factors that play a role in determining what premiums companies can charge for insurance coverage. State regulations often limit insurers in their rate-setting process, and competition may play a role in what they may be able to charge in the marketplace. In the ensuing discussion, however, we assume that insurers are free to set the premiums at any level they wish. Even in the absence of regulation and competition, there are two major problems that insurers face when deciding to provide coverage against earthquakes: ambiguity and highly correlated risks. Turning rst to ambiguity, Kunreuther et al. [1995] conducted a survey of 896 underwriters in 190 randomly chosen insurance companies to determine what premiums would be required to insure a factory against property damage from a severe earthquake. The surveys goal was to examine changes in pricing strategy as a function of the degree of uncertainty in the expected loss. Because the expected loss, E[L], of any event is dened as the probability of the event occurring multiplied by the resulting loss, probability and loss were the two variables considered. A probability is well specied when there is enough historical information on the event such that all experts agree that the probability of a given loss is p. When there is wide disagreement about the estimate of p among the experts, this ambiguous probability is referred to as AP. Similarly, L represents a known loss (i.e., there is a general consensus about what the loss will be if a specic event occurs). When a loss is uncertain, the experts estimates range between Lmin and Lmax, and this uncertain loss is denoted UL. Combining the degree of probability and loss uncertainty leads to four cases, which are shown in Table 32.1, along with a set of illustrative examples of the types of risks that fall in each category. The case of an earthquake falls within case 4, where the probability is ambiguous, Ap, and the loss is unknown, UL. As a further part of the survey, in order to see how underwriters might set premiums based on degrees of ambiguity and uncertainty, four scenarios were constructed for each participant to consider, as shown in Table 32.1. When the risk is well specied, the annual probability of the earthquake is either 1.0 or 0.5%; the loss, should the event occur, is either $1 million or $10 million. With the underwriters premium standardized to one for the nonambiguous case (i.e., p is well specied and L is known), one can examine how ambiguity affects pricing decisions.
TABLE 32.1 Classication of Risks by Degree of Ambiguity and Uncertainty
Loss Probability Well-specied Known Case 1 p, L (life, auto, re) Case 2 Ap, L (satellite) Unknown Case 3 p, UL (playground accidents) Case 4 Ap, UL (earthquake, bioterrorism)

Ambiguous

2003 by CRC Press LLC

32-6

Earthquake Engineering Handbook

TABLE 32.2 Ratios of Underwriters Actuarial Premiums for Ambiguous and Uncertain Earthquake Risks Relative to Well-Specied Risks
Cases Scenario p = 0.5% L = $1 million E[L] = $5,000 p = 0.5% L = $10 million E[L] = $50,000 p = 1% L = $1 million E[L] = $10,000 p = 1% L = $10 million E[L] = $100,000 Case 1 (p, L) 1 Case 2 (Ap, L) 1.28 Case 3 (p, UL) 1.19 Case 4 (Ap, UL) 1.77

1.31

1.29

1.59

1.19

1.21

1.50

1.38

1.15

1.43

From Table 32.2, the pricing ratios of the other three cases relative to the nonambiguous case are presented. For the highly ambiguous case (i.e., Ap and UL), the premiums were between 1.43 to 1.77 times higher than if underwriters priced a nonambiguous risk. As expected, the pricing ratios for the other two cases were always above one but less than the highly ambiguous case. Correlated risk is also an important factor for insurers to examine in determining what premium to charge to make a risk insurable. Correlated risk refers to the simultaneous occurrence of many losses from a single event. As pointed out earlier, earthquakes produce highly correlated losses; many homes and buildings in the affected area are damaged and destroyed by a single event. If a risk-averse insurer faces highly correlated losses from one event, it may want to set a high enough premium not only to cover its expected losses but also to protect itself against the possibility of experiencing earthquake losses. An insurer will face this problem if it has many policies in one area, such as providing earthquake coverage to residences only in Los Angeles County rather than diversifying across the entire western region of the United States. To illustrate the impact of correlated risk on the distribution of losses, assume that there are two policies sold against a risk where p = 10% and L = $100. The expected loss for each policy is $10. If the losses are perfectly correlated, there will be either two losses with a probability of 10%, or no loss with a probability of 90%. On the other hand, if the losses are independent of each other, the chance of two losses decreases to 1% (i.e., 0.l * 0.1), with the probability of no loss equal to 81% (i.e., 0.9 * 0.9). In this case, there is also an 18% chance that there will be only one loss (i.e., 0.9 * 0.1 0.1 * 0.9). The expected loss for both the correlated and uncorrelated risks is $20.2 However, the variance will always be higher for the correlated risk. Thus, risk-averse insurers will always want to charge a higher premium for the correlated risk. There are two other problems insurers face in setting premiums with respect to risks: adverse selection and moral hazard. Neither appears to be a major problem with respect to earthquake risks. Adverse selection occurs when the insurer cannot distinguish between the probability of a loss for good and poor risk categories. Moral hazard refers to an increase in the probability of loss or the magnitude of the loss itself caused by the behavior of the policyholder. With regard to earthquakes, adverse selection has been a problem for insurers until recently. Insurers can normally identify houses that are low and high risks with respect to potential losses from disasters but had not done so until the 1980s. Even then, several companies were rudely surprised by their losses

For the correlated risk, the expected loss is 0.9 * $0 + 0.1 * $200 = $20. For the independent risk, the expected loss is (0.81 * $0) + (0.18 * $100) + (0.01 * $200) = $20.
2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-7

$100

Value

Layer C ($56 p/o $70 xs $30)

Part D
(retained)

80%
$30

Layer B ($20 xs $10)


$10 0

Layer A (typ. Deductible, eg, 10%) Fraction

FIGURE 32.2 Insurance diagram.

in the 1994 Northridge earthquake. If properties are priced to reect risk, there cannot be adverse selection. Even if the property is located in an area of high seismic activity or on poor soil, if the earthquake insurance is priced properly, there cannot be adverse selection. However, pricing properties properly implies perfect knowledge, which is never possible, so that adverse selection in practice requires diligence on the part of insurers. An example of moral hazard is setting re to a house that has been damaged by an earthquake; the house is insured for re but may not be insured for earthquake, and thus the homeowner collects on the re policy. Insurers do not like ambiguity and this is reected in the premiums that they charge for events that fall into this category. Not surprisingly, the higher the uncertainty regarding the probability of a specic loss and its magnitude, the higher the premium will be. As shown by a series of empirical studies, actuaries and underwriters are generally averse to ambiguity and they tend to recommend much higher premiums for risks with high degrees of uncertainty surrounding them.

32.2.3 Types of Insurance


There are many types of insurance, the broadest distinction being layered vs. coinsurance. This is illustrated in Figure 32.2, which is termed an insurance diagram. On the right is the pdf of loss for the property under consideration (note that the pdf in Figure 32.2 is turned on its side, vs. how the pdf is shown in Figure 32.1), while the large rectangle on the left illustrates the way the coverage is divided between the different types of insurance. On the bottom, layer A represents the risk up to the rst 10% of loss of the value of the property because this example shows the total insured value (TIV) as $100, the deductible is 10% of $100 or $10. If the property sustains a $9 loss, the insurer of layer A sustains all of that $9 loss; if the property sustains a $12 loss, the insurer of layer A sustains a loss equal to 10% of the value of the property (i.e., $10), and the insurer of layer B sustains the next $2 of loss. Layer A is the rst layer, and is typical of the deductible on ordinary building properties (in the case of the deductible, the insurer of layer A is the owner, who retains the risk). Layer B is the second layer, and represents a risk of 20% of the value in excess of the rst 10%. Because the value of the property in this example is $100, this layer would be denoted $20 xs $10. There is no coinsurance in layer B. Layer C is divided (or has coinsurance) such that the insurer bears 80% of the risk (denoted $56 part of [p/o] $70 xs $30), while the owner bears (or retains) the remaining 14% of the value (denoted as Part D). Note that the higher the layer (for the pdf shown), the lower the probability of loss, so that the required premium (per unit value) for higher layers should be relatively lower.

32.2.4 Reinsurance
Reinsurance is a transaction whereby the reinsurer, for a consideration, agrees to indemnify the ceding company (the primary insurer) against all or part of the loss which the insurer may sustain under the
2003 by CRC Press LLC

32-8

Earthquake Engineering Handbook

policy or policies it has issued. Basically, reinsurance is insurance for an insurer. Reinsurance is a very common practice and is employed, analogously to insurance, to permit the ceding company to spread its risk, and thereby better manage its total exposure and its cash ow. Reinsurance has two basic forms: pro rata and excess of loss. Pro rata reinsurance is sometimes called proportional reinsurance (analogous to coinsurance), because the reinsurer shares proportionately in risks and premiums. Pro rata reinsurance is often contracted on a treaty basis, such that it occurs automatically, where the reinsurer agrees to assume some part (e.g., 20% of all claims) of the ceding companys risks of a predened nature (e.g., all homeowner re policies). In effect, the reinsurer becomes a nancial partner of the primary company, lending its capital to support the primary insurers capital. In contrast, excess of loss (analogous to layer insurance) reinsurance provides for loss indemnication above a specied level (attachment point) up to a limit (exhaustion point), which denes a layer. The reinsurer will charge a premium for losses that occur within the layer ceded. Alternative to automatic reinsurance is facultative reinsurance where, case by case, a primary insurer transfers some portion of a specic risk to a reinsurer. A small reinsurer may have the opportunity to underwrite an unusually large risk, which its capital surplus would not normally permit it to insure. By ceding or laying off a major portion of the large risk to a reinsurer, the ceding company utilizes the reinsurers capital to appear to underwrite or assume the large risk, thus improving client relations, reputation, and (hopefully) prots. Determination of reinsurance risk and premiums is performed analogously to ordinary insurance, and diagrams similar to Figure 32.2 are developed for a ceding companys portfolio, to determine the actuarial basis for the loss potential and associated premium. The development of the pdf of the ceding companys portfolio risk is a specialized eld, performed by only a few modeling companies. As noted previously, any insurer assuming a risk from another insurer is technically a reinsurer, and many primary companies have a reinsurance operation. Globally, however, reinsurance is dominated by a relatively few companies with access to the necessary capital. Traditional global reinsurers include several Lloyds syndicates, Munich Re, Swiss Re, SCOR, and in the United States, Gen Re and Employers Re. In the 1990s, a major and growing market for reinsurance also developed in Bermuda, competitive with the traditional reinsurers. Analyzing a ceding companys reinsurance needs and efciently accessing the reinsurance market is a complex task, such that several large reinsurance intermediaries play a major role, including Guy Carpenter (part of the Marsh McLennan group), Aon, and Beneld Blanche. Reinsurance plays a very signicant role with respect to natural hazards catastrophic risks, such as earthquake. As noted previously, a prime distinguishing feature of large earthquakes is their simultaneous impact on thousands of properties, resulting in loss of independence between risks. The net result is that the losses due to large cats is far more uncertain than ordinary insurance risks, so that many primary companies utilize reinsurance to a signicant degree in managing catastrophe risk. This reliance on reinsurance extends to large regional catastrophe pools, such as the California Earthquake Authority or New Zealands Earthquake Commission, which employ reinsurance in multibillion dollar programs of several layers.

32.3 Earthquake Insurance


In the United States, earthquake insurance is typically offered to homeowners as an amendment to their re or homeowners policy, and to commercial customers as an added peril or difference in condition (DIC) to whichever type of loss (i.e., re, BI, etc.) they are concerned about. One important point should be noted: in the United States, the United Kingdom, Canada, Australia, New Zealand, and other Englishspeaking countries, re following earthquake is covered under the re policy, whereas in virtually the rest of the world, re following earthquake is covered under the earthquake policy. Because conagration following earthquake is a signicant risk in regions with a large stock of wood buildings, such as California, New Zealand, and Japan, this distinction can be critical. Fire following earthquake is addressed in Chapter 29 of this volume.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-9

$60

Economic Losses
($ billions)

Median
0

Median Loss

Earthquake (5 events) Flood (21 events) Hurricane (25 events)

Type of Natural Disaster

FIGURE 32.3 Historical economic losses vs. type of signicant U.S. natural hazards in 50-year period. (Courtesy Geoscience Division, American Re)

A considerable amount of historical data is utilized to estimate earthquake insurance premiums as a function of the type of building (e.g., steel vs. masonry). Premiums are fundamentally based on two factors: frequency (how often an event such as an earthquake occurs), and severity (what is the loss, when it occurs). With respect to earthquakes, given the low frequency of these events over time, historical data on which to base insurance premiums are limited. Hence, there is a need to supplement historical information with scientic data in order to improve the accuracy of the risk estimates and determine what premiums to charge. Because there may be considerable uncertainty surrounding these gures, insurers may charge relatively high premiums, reecting their aversion to ambiguity and uncertainty. The important question is whether or not there is sufcient demand for coverage at the rates that insurers want to charge. Earthquakes pose a set of challenging problems for the insurance industry because they involve potentially high losses that are extremely uncertain. This ambiguity is best seen in a graphical representation of historical loss data. Figure 32.3 depicts the range of economic losses in the United States for three types of events: earthquakes, oods, and hurricanes. The natural hazards shown have corresponding frequencies of occurrence of 5, 21, and 25 events, respectively, over the past 50 years in the United States. The horizontal axis indicates the type of natural disaster3 and the vertical axis shows the range of economic loss and median value of loss. These natural hazards have a historically low median and a signicant upper range of economic loss. In other words, while the historical losses have been relatively low for more than half of these disasters as shown by the median lines in Figure 32.3, the upper bound of the range of losses has been very high. More interestingly, the most infrequent type of natural disaster (earthquakes) and the most frequent type of natural disaster (hurricanes) have the greatest ranges of historical loss. Given this signicant uncertainty, it is not surprising that earthquake models were initially developed to aid the insurance industry in the quantication of potential loss from these two types of disasters.

32.3.1 Insurance Supply and Insurers Decision Processes


The above discussion suggests that, in theory, insurers can offer protection against any risk that they can identify and calculate the expected losses for, as long as they have the freedom to set premiums at any level. However, due to problems of ambiguity and highly correlated losses, they may want to charge premiums that considerably exceed the expected loss. For some risks, the desired premium may be so
3

A signicant U.S. natural disaster is an economic loss (adjusted to year 2000 dollars) of at least $1 billion or more than 50 deaths, or both, attributed to the event. Also, these loss estimates, while adjusted for ination, are not adjusted for population increases and increases in wealth in given locations.
2003 by CRC Press LLC

32-10

Earthquake Engineering Handbook

high that there would be very little demand for coverage at that rate. In such cases, even though an insurer determines that a particular risk meets the two insurability conditions discussed previously, it will not invest the time and money to develop the product because it believes that it cannot protably market it. More specically, the insurer must be convinced that there is sufcient demand to cover the development and marketing costs of the coverage through future premiums received.4 In addition, insurers must be convinced that they can manage the ambiguity of the risk so that premiums can be charged that will be attractive to potential policyholders. Insurers will limit their losses by restricting the amount of coverage in hazard-prone areas to avoid losses that may cause them to lose a considerable amount of surplus and possibly become insolvent. In his denitive study, which sheds light on insurers and their decision rules, Stone [1973] indicated that rms interested in maximizing expected prots are subject to two constraints representing the survival of the rm and the stability of its operation. The insurance underwriter satises the survival constraint by choosing a portfolio of risks with an overall expected probability of insolvency less than some threshold, p1 . The stability constraint focuses on the combined loss and expense ratio, LR, for each year. Insurers dene a target level, LR*, which represents an upper limit on this ratio and requires that the probability that LR exceeds LR* is less than p2. A simple example illustrates how an insurer, in determining whether the earthquake risk is insurable, will utilize these two constraints. Assume that all homes in an earthquake-prone area are equally resistant to damage, so that the insurance premium on each structure is the same, denoted P. Further assume that the Down-to-Earth (DTE) Insurance Company has $A dollars in current surplus and wants to determine the number of policies it can write and still satisfy its survival and stability constraints. Then, the maximum number of policies, n1 , satisfying the survival constraint is: Probability [(Total Losses Expenses n1P A)] p1 In addition, the maximum number of policies, n2, satisfying the stability constraint is: Probability [(Total Losses Expenses)/n2P LR*] p2 (32.4) (32.3)

Whether DTE will view the earthquake risk as insurable depends on whether the xed cost of issuing policies in the area is sufciently low so it can make a positive expected prot. This, in turn, depends on how large the values of n1 and n2 are for any given premium, P. Note that DTE also has some freedom to change its premium. A larger P will increase the values of n1 and n2 but will lower the demand for coverage. DTE will decide not to offer earthquake coverage if it believes it cannot attract enough demand at any premium structure to make a positive expected prot using the survival and stability constraints as restrictions on how many policies it is willing to offer. In a series of interviews conducted with insurers following Hurricane Andrew and the Northridge earthquake, it was indicated that a key factor that inuences insurers decisions on how many policies to write is their probable maximum loss (PML) in the event a disaster occurs.5 Today, in many hazardprone regions, most insurers would like to reduce their current PML estimates. A.M. Best Company includes PML exposures as a part of its rating of insurer capability [BestWeek, 1996].

32.3.2 Insurance Demand and Homeowners Decision Processes


Given that insurers provide residential coverage for earthquakes, how much interest does the potential policyholder have in purchasing this coverage at some prespecied price? The uncertainty associated with

We are assuming implicitly that there are no regulatory restrictions that force the insurer to keep the premium for certain types of coverage below a given value. If that value is too low, the insurer may decide not to offer such coverage simply because it will not be protable. 5 These interviews were conducted by Jacqueline Meszaros as part of a National Science Foundation study, The Role of Insurance and Regulations in Managing Catastrophic Risk.
2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-11

earthquakes, coupled with the limited information-processing capabilities of individuals, lead homeowners to utilize simplied decision rules in determining whether or not they want to purchase coverage.6 Residents in hazard-prone areas often exhibit one of two reactions with respect to earthquake events. If they have not experienced a specic disaster and do not have friends and neighbors who have been affected by such an event, they often behave as if it will not happen to me. In other words, they treat the probability of an earthquake as essentially zero. These residents will have no interest in voluntarily purchasing insurance [Kunreuther, 1996]. Alternatively, individuals who have recently experienced a disaster are much more likely to purchase insurance voluntarily. With the recent event prominent in their minds, the perceived probability of occurrence triggers interest in protecting themselves. It is thus not surprising that demand for earthquake insurance in California increased signicantly following the Loma Prieta earthquake in 1989 and the Northridge earthquake in 1994; in the 1970s, less than 10% of homes in the state were insured against earthquake damage. Similar behavior occurred in Japan following the 1995 Kobe earthquake. The individuals who are not concerned with the consequences of a future disaster often behave as if they have a threshold probability, p*, below which they will not purchase insurance. In their minds, the attention costs (i.e., the costs to think about the event) and the transaction costs (i.e., the costs to gather information about insurance coverage) are too high relative to the probability that the disaster will occur. The higher these perceived costs, the larger p* will be. Those who feel the probability of a disaster exceeds p* are likely to seek out friends and neighbors for information about where to buy coverage [Kunreuther et al., 1978; Weinstein, 1987].

32.4 Earthquake Insurance Risk Assessment


The ambiguity and uncertainty associated with expected losses from earthquakes highlight the importance of understanding the risk assessment process and how it can potentially reduce the uncertainty associated with these earthquake events. The risk assessment process is most commonly probabilistic in nature, consisting of two primary components, hazard and vulnerability, both of which are covered in considerable detail in other chapters in this volume, and thus only summarized here. This section rst reviews traditional methods insurers used to estimate their overall risk, and then discusses earthquake loss modeling, which built on earlier work [Hirschberg et al., 1978; Petak and Atkinsson, 1982; Scawthorn, 1981; Steinbrugge, 1982; Wiggins, 1981] to emerge in the 1990s as a powerful tool for the insurance industry, leading to innovations, such as catastrophe securitization (discussed later).

32.4.1 California Department of Insurance Probable Maximum Loss


In the aftermath of the San Fernando earthquake in 1971, concern about the exposure of the insurance industry to earthquakes greatly increased, leading the California Insurance Department to issue Rule 226, which requires all licensed insurers to report each year their insured exposures for earthquake shake damage on residential and commercial structures in California. At that time, the percentage of homes and commercial structures insured for earthquake damage was less than 10%, and the insurance losses from the San Fernando earthquake were about $46 million. Since then, the demand for earthquake insurance has grown dramatically, along with a dramatic increase in housing and commercial building values. The insurance industry paid out about $1 billion after the Loma Prieta earthquake in northern California in 1989. The insurance industry has paid out more than $15 billion for the January 17, 1994 Northridge earthquake. While no insurers went out of business after the Northridge earthquake, the nancial impact was severely constricting for insurers of all sizes.

6 These decision rules differ from normative models of choice, such as expected utility theory or costbenet analysis. For more detail on how they differ, see Camerer and Kunreuther [1989].

2003 by CRC Press LLC

32-12

Earthquake Engineering Handbook

As a result of Rule 226, the California Insurance Department developed a simple but useful method for insurers to estimate their probable maximum loss (PML). Denitions of PML vary widely within the industry but the methodology developed by the California Department of Insurance and its consultant K.V. Steinbrugge has been used by the State of California since about 1980 to monitor insurance industry exposure [Steinbrugge, 1982; Steinbrugge and Roth, 1994; Roth and Van, 1998; CDI, 2001]. In that methodology: Building Class PML (i.e., for an individual building of a specic class, such as wood frame, see Tables 32.3 and 32.4) is dened as the expected maximum percentage of monetary loss which will not be exceeded for nine out of ten buildings, where the building is located on rm alluvial ground, subjected only to the vibratory motion from the maximum probable earthquake (i.e., not astride a fault or in a resulting landslide). Aggregate PML is the sum of all of the PML values in a PML zone (see Figure 32.4), plus factored PML values for buildings located outside of the PML zone but still within the earthquake underwriting zone. A factored PML is a reduced PML value based on reduced intensity (i.e., damage) with increasing distance away from the causative fault. Using this methodology, insurance companies in California are required to report their aggregrate PML each year to the California Department of Insurance [CDI, 2001]. This is of interest to the department, as it wishes to assure adequate company surplus to assure payment of claims in the event of a large earthquake.

32.4.2 CRESTA Zonation


A key system employed by reinsurers for tracking catastrophe exposure is the CRESTA (Catastrophe Risk Evaluating and Standardizing Target Accumulations) system,7 originally developed in the late 1970s in a joint project of several major reinsurance companies. CRESTA is a standardized method of exposure accumulation reporting by primary companies for reporting to reinsurance companies. Based primarily on the observed or expected seismic activity within a country, CRESTA zones permit accumulation of insured values within a country by administrative or political boundaries for easier assessment of risks. It is a simple but effective system that has permitted rational albeit approximate analysis of the potential exposure by the large, multinational reinsurance companies. Figure 32.5 shows examples of CRESTA accumulation zones; note that the California CRESTA zones are identical to the California Department of Insurances earthquake zones, which is typical in that the CRESTA system seeks to minimize the reporting burden on companies by using existing reporting formats wherever possible. CRESTA reports, for example, are utilized by a reinsurer to compare requests for reinsurance from two primary companies. The two primaries may be requesting catastrophe reinsurance for properties located in several countries. Such global programs can become very complex and difcult to analyze.
TABLE 32.3 Building Classes, California Department of Insurance
Class 1A 1B 1C Type of Building Single- through four-family dwellings. No limitations on story height, area, and construction materials. Homeowners. Habitational: Wood frame and frame stucco habitational buildings that do not exceed two stories in height, regardless of area. Nonhabitational: Wood frame and frame stucco, except buildings: (1) > three stories and (2) > 3000 ft 2 in ground oor area. Wood frame and frame stucco buildings not qualifying under Class 1C. Mobile homes and contents.

1D 1E

The systems name actually derives from the fact that the initial meeting was held at the Hotel Cresta.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-13

TABLE 32.3 (CONTINUED) Building Classes, California Department of Insurance


All-Metal Buildings 2A 2B All-metal buildings one story in height and 20,000 ft2 or less in ground oor area. Wood or cement-asbestos are acceptable alternatives to metal roong and siding. Buildings that would qualify as Class 2A, except for exceeding area or height limitations. Steel Frame Buildings 3A Buildings with a complete steel frame carrying all loads. Floors and roofs must be of cast-in-place (CIP) reinforced concrete (RC) or of concrete ll on metal decking welded to the steel frame (open web steel joists excluded). Exterior walls must be nonload bearing and of CIP RC or of reinforced unit masonry. Buildings having columnfree areas greater than 2500 ft2 (auditoriums, theaters, public halls, etc.) do not qualify. Buildings with a complete steel frame carrying all loads. Floors and roofs must be of CIP RC, metal, or any combination thereof, except that roofs on buildings over three stories may be of any material. Exterior and interior walls may be of any nonload-bearing material. Buildings having a complete steel frame with oors and roofs of any material (such as wood joist on steel beams) and with walls of any nonload-bearing materials. RC Buildings Combined RC and Structural Steel Buildings Note Class 4A and 4B buildings must have all vertical loads carried by a structural system consisting of one or a combination of the following: (1) CIP RC frame, (2) CIP RC bearing walls, (3) partial structural steel frame with (1) or (2) or both (1) and (2). Floors and roofs must be of CIP RC, except that materials other than RC may be used for the roofs of buildings over three stories. Buildings with a structural system as dened by the note above with CIP RC exterior walls or reinforced unit masonry exterior walls. Not qualifying are buildings having column-free areas greater than 2500 ft2 (auditoriums, theaters, public halls, etc.). Buildings having a structural system as dened by the note above with exterior and interior nonbearing walls of any material. Buildings having (1) a partial or complete load carrying system of precast concrete or (2) RC lift-slab oors or roofs, or both (1) and (2), and (3) otherwise qualifying for Class 4A and 4B. Buildings having an RC frame or combined RC and structural steel frame. Floors and roofs may be of any material (such as wood joist on RC beams), walls may be of any nonload-bearing material. Mixed Construction 5A Buildings having load-bearing exterior walls of (1) CIP RC or (2) precast RC (such as tilt-up walls) or (3) reinforced brick masonry, or (4) reinforced hollow concrete block masonry or (5) any combination of (1)(4). Floors and roofs may be of wood, metal, CIP concrete, precast concrete, or other material. Interior bearing walls must be of wood frame or any one of a combination of the aforementioned wall materials. Note: No class distinction is made between newer, highly earthquake-resistive buildings and older moderate earthquake-resistive buildings having these construction materials. ISO Classes 5A and 5AA shall be combined and considered as Class 5A. Buildings having load-bearing walls of unreinforced brick or other types of unreinforced solid unit masonry, excluding adobe. Buildings having load-bearing walls of hollow tile or other hollow unit masonry construction, adobe, and cavity wall construction. Also included are buildings not covered by any other class. Type of Building Earthquake-Resistive Construction 6 Any building with any combination of materials so designed and constructed as to be highly earthquake resistant and also with superior damage control features in addition to the minimum requirements of building codes. Miscellaneous 7 Bridges, tunnels, dams, piers, wharves, tanks, tank contents, towers of all types, and the like. Time-element coverages for these structures to be included. Source: http://www.insurance.ca.gov/RRD/RSU/Forms/Year2001/PML2001/pml2001.htm.
2003 by CRC Press LLC

3B

3C

4A

4B 4C 4D

5B 5C Class

32-14

Earthquake Engineering Handbook

TABLE 32.4 Probable Maximum Loss (PML) for Building Classes, California Department of Insurance
CONSTRUCTION CLASSES, PML, AND DEDUCTIBLES Net PML (%) Class 1A & 1B Deductible 1% 5% 10% 15% Mini Wrap 1C 1D 1E 2A 2B 3A 3B 3C 4A 4B 4C 4D 5A 5B 5C 6 *7 COC 5% 5% 2% 5% 5% 5% 5% 10% 5% 5% 10% 10% 5% 10% 10% 5% 0% ** Zone A 6.75 3.63 2.13 1.38 0.69 2.94 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone B 5.75 3.00 1.63 1.00 0.50 2.50 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone C 6.13 3.13 1.75 1.13 0.56 2.56 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone D 2.63 1.19 0.56 0.31 0.16 1.03 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone E 5.25 2.38 1.13 0.63 0.31 2.06 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone F 3.13 1.88 1.13 0.63 0.31 1.56 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone G 1.75 1.00 0.63 0.38 0.19 0.81 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 ** Zone H 2.50 1.50 0.88 0.50 0.25 1.25 3 10 5 2 10 15 25 25 20 35 50 45 25 60 75 10 50 **

* Includes special structures such as bridges, tunnels, dams, piers, wharves, tanks, tank contents, towers of all types, and the like. Timeelement coverages for these structures are also to be included. **Buildings in the course of construction (COC) are to be placed in the completed building class, using 50% of the completed PML and the full value of the usual deductible (Fire Forms where insurance is written at 80% of value or higher). Buildings constructed of materials of more than one class shall be assigned to the Construction Class with the highest PML. Earthquake liabilities on buildings, contents, time element, and other location coverages shall be included under the building Construction Class. Source: http://www.insurance.ca.gov/RRD/RSU/Forms/Year2001/PML2001/pml2001.htm.

Because the two companies report in a standardized CRESTA format, the aggregate risks for the two companies can be compared on the same basis to allow the reinsurer to make appropriate pricing and other decisions. CRESTA was developed prior to the emergence of modeling for the global insurance industry, and is gradually being supplanted by probabilistic risk assessment but will continue to be useful for a signicant period as a simple standardized reporting format.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-15

ZONE
Del Norte Siskiyou Modoc

COUNTY Alameda Contra Costa Del Norte Humboldt Lake Marin Mendocino Monterey Napa San Benito San Francisco San Mateo Santa Clara Santa Cruz Solano Sonoma Los Angeles Orange Kern San Luis Obispo Santa Barbara Ventura San Diego

ZONE E

COUNTY Alpine Imperial Inyo Mono Riverside San Bernardino Fresno Kings Madera Mariposa Merced Tulare Amador Butte Calaveras Colusa El Dorado Glenn Nevada Placer Sacramento San Joaquin Stanislaus Sutter Tuolumne Yolo Yuba Lassen Modoc Plumas Shasta Sierra Siskiyou Tehama Trinity

Humboldt

Trinity

Shasta

H
Lassen Plumas Butte

A
Mendocino

Tehama

Glenn

Sierra Nevada

B C

Sutte

Lake

Colusa

Yuba

Placer

Yolo Sonoma Napa Solano Contra Costa

G
me nto cra Sa
San Joaquin

El Dorado Alpine Amador Calaveras Tuolumne Mono

A
Marin San Francisco

San Mateo

Alameda

Stanislaus Merced

Mariposa

E
Inyo

Santa Clara Sa nta Cr uz San Benito

Madera

Fresno

Monterey Kings

Tulare

San Luis Obispo

Kern

C
Santa Barbara Ventura Los Angeles

San Bernardino

B
Orange

E
Riverside

Imperial

San Diego

FIGURE 32.4 California Department of Insurance earthquake zones. Source: http://www.insurance.ca.gov/RRD/ RSU/Forms/Year2001/PML2001/pml2001.htm)

32.4.3 Probabilistic Earthquake Insurance Loss Estimation


The CRESTA system was critically important for insurers8 and reinsurers when these companies did not have the ability to accurately estimate the potential losses a catastrophe may cause to the properties they insure. By the 1990s, however, and building on earlier work [Hirschberg et al., 1978; Petak and Atkinsson,

8 Especially for large global primary companies, which used the system internally as well as for reporting to reinsurers.

2003 by CRC Press LLC

32-16

Earthquake Engineering Handbook

Earthquake Accumulation Assessment Zones

California, USA

6 8

H
12

3 2 1

13

10 9 8 8 14 11 7 15 6 5 9 10 11 4 12 San 4 3 Francisco 3 2 2 6 13 14 15 16

1 1 5 4 3 2 1 2 3 2

Fresno

C
2

E
4

B
1 3 5

Los Angeles
0 100 200 miles

D
San Diego

1
Detailed map of Zone
SAITAMA IBARAKI

HOKKAIDO

Tokyo 5.2 Chiba YAMANASHI 5.3.3 5.1 KANAGAWA CHIBA SHIZUOKA

Sea of Japan Pacific Ocean

2.1

0SHIMA

Kawasaki 5.3.1 Yokohama 5.3.2

2.2

3.1

2.3 3.2

40 km
2.4 3.3

3
4.2 7.1 4.1 7.2 4.4 6.2 4.3 7.3 6.1 6.3 6.5 6.4

HONSHU

9
9.4 9.1 9.2 9.3 9.5

4 5

8.1 8.28.3 8.4 6.6 8.5 8.6

See detailed map

11
11

10

10
SHIKOKU
0

OKINAWA

KYUSHU

100 200 300 km

Naha

12
0 25 50 km

FIGURE 32.5 CRESTA Zones for California and Japan.


2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-17

Hazard Source and Attenuation Models

Hazard Module

Historic and Scientific Hazard Information

Structure, Content, and BI Vulnerability Models

Damage Module

Inventory of Asset Attributes

Loss Module

Loss inuring Information

FIGURE 32.6 Basic earthquake insurance loss estimation process.

1982; Scawthorn, 1981; Steinbrugge, 1982; Wiggins, 1981], the technology of earthquake loss modeling emerged as a powerful tool for the insurance industry. Earthquake loss modeling has applications in many areas, such as infrastructure planning and building-code development, and is covered in detail elsewhere in this volume. The discussion here is an overview only, limited to an insurance perspective. The basic earthquake insurance loss estimation process is shown in Figure 32.6, and begins with the hazard component, which calculates the probability of exceeding certain levels of event intensity based on the physical parameters that describe the earthquake. The vulnerability component deals with the nature of the damage to the built environment (e.g., structures and lifelines) and the resulting monetary cost of repairs or other losses. Due to the complexity of the hazard and the limited historical data on past events, both of these components have considerable uncertainty. When the potential hazard is determined, one needs to characterize the population and property at risk. One way to specify who is at risk is to construct a community or region consisting of homes, businesses, and other properties that are subject to future disasters. One needs to know the design of each structure, whether specic loss reduction measures are in place or can be utilized in the future, and the propertys location in relation to the hazard (e.g., distance from an earthquake fault line), as well as other risk-related factors. Using the estimates of probability that earthquakes of certain intensities or magnitudes will occur, the expected damage to a community or region (or portfolio of structures) is calculated. From this estimate, the resulting loss to the structures is estimated. The recent use of geographic information systems (GIS) for incorporating geologic, topographical, and property information for a region has enabled scientists to estimate potential damage and losses from different types of hazards much more easily. The data for the region are stored in the form of GIS maps of hazard estimation, maps of secondary hazards, and maps of damage to structures in the region [King and Kiremidjian, 1997]. These maps form the core of todays earthquake models. A key tool utilized when undertaking an analysis for determining potential earthquake risk is the loss exceedance probability (EP) curve. The exceedance probability of loss (i.e., the probability of exceeding the loss) is the complement of the cumulative distribution function (CDF) of loss: EP (L) = 1 CDF (L) (32.5)

The use of loss exceedance curves has enabled the insurance industry to develop more accurate rate structures for coverage against losses from earthquakes. An example of an exceedance probability curve is shown in Figure 32.7. Typically, the exceedance probability curve has a continuous shape with the probability of exceedance on the y-axis and the measure of risk on the x-axis. The probability of exceedance is the probability that the measure of risk will be surpassed for a given period of time. Most often, the period of time is 1 year (i.e., an annual probability of exceedance).
2003 by CRC Press LLC

32-18

Earthquake Engineering Handbook

Probability of Exceedance, p

Loss, L (in Dollars)

PML

FIGURE 32.7 Exceedance probability curve.

Loss, L (in Dollars)

$20

10%

$20

$10

15%

$10

80%

Probability of exceedance, p

FIGURE 32.8 Examples: (left) insurance diagram and (right) loss exceedance probability curve with layer.

The measure of risk (x-axis) in Figure 32.7 is dollar loss. As expected, the probability of exceedance decreases as the risk measure increases. The events with the lowest probabilities of exceedance are the ones with the highest potential loss; typically, these are the events of greatest concern. As previously mentioned, developing exceedance probability curves for earthquakes has signicant uncertainty, which plays a critical role in assessing natural disaster mitigation and nancing the consequences of earthquakes. It is important to understand where the uncertainty lies, and how it can be captured and utilized in the risk assessment and risk management process. One way to capture this uncertainty is through a set of exceedance probability curves. In other words, values are calculated at various condence intervals around the mean estimates of loss (or hazard or damage). Given the loss exceedance probability curve and an insurance diagram, the probability of sustaining a loss, per an insurance contract, can be readily determined. For example, referring to Figure 32.8), the insurance contract in this case is for a particular layer, specically $10 excess of $10 (which could easily be $10 million but $10 is used as an example; note that this is a layer with an upper limit of $20 and a lower limit of $10, written 10 xs 10). Furthermore, as is typical in many contracts, the insurer will only cover 80% of this layer, requiring the insured to retain 20% (in part to limit the potential for moral hazard). If the property in this example sustains a loss of $8, for example, the insurer pays nothing. If the loss is $15, the insurer pays 80% of the excess of $10, or $4. If the loss is $25, the insurer pays 80% of the $10 xs $10, or $8. The maximum the insurer may pay is $8, no matter how high the loss. From the exceedance probability curve, we note that the probability of penetrating the layer is 15%, and of exhausting the layer is 10%. Therefore, the expected loss given these parameters is approximately
2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-19

capital

failure time

FIGURE 32.9 Example of dynamic nancial analysis.

80% of (10% of $10 plus half of 5% $10), or 80% of $1.25, or $1. This is the pure premium, given the hazard, vulnerability, values, and insurance contract. One dollar, or 10%, is therefore the rate on line for 80% coverage of $10 xs $10. Note that the insurance contracts terms are a major determinant in the insurers expected loss, and that a high layer on a risky property can still have a low expected loss, and therefore may be a protable contract for the insurer while still providing some protection to the insured.

32.4.4 Dynamic Financial Analysis


An emerging technique for insurers and insureds is dynamic nancial analysis (DFA), which is simply a forward-looking estimate of the nancial condition of a risk or risk pool over a period of time (e.g., 50 years). DFA models are employed (1) by insurance companies, to calculate the capital required for an insurance pool, given a desired level of solvency;9 and (2) by insureds, to estimate what their risk is, and how it should be priced (i.e., to determine if the price available from insurers is a good buy). In actuality, DFA is a tool currently being employed only by larger, more progressive insurers, and by only the largest, most sophisticated insureds, such as multinational companies or government pools seeking reinsurance. The explanation of a DFA model provided here is a somewhat simplied version of a relatively complex model typically used in insurance company studies (for a detailed description of DFA modeling, see the DFA Research Handbook, Casualty Actuarial Society, http://www.casact.org/research/dfa/). Conceptually, a DFA model has three steps: 1. A set amount of capital is invested at the initiation of the insurance pool. 2. Premiums are collected, invested annually, with return on the investment. 3. The insurance pool pays out losses of claims for simulated events, nances mitigation efforts, and covers its own expenses. In order to estimate annual risk losses to the insurance pool, Monte Carlo simulations are employed to generate loss samples consistent with the loss exceedance distributions calculated as described previously. Each loss sample has an equal likelihood of occurrence. Because the annual aggregate loss distributions are used, the occurrence of more than one earthquake in a given year is accounted for. Schematically, the process is shown in Figure 32.9, where each simulation begins with an initial capital, and undergoes 50 years of premium and investment growth and risk claim losses. In most simulations, the pool performs well but occasionally an infelicitous combination of exposure growth and risk losses leads to exhaustion of capital, or failure (i.e., bankruptcy). Based on the risk loss sample, a random walk technique is used to generate 10,000 trials (or random walks) of 50 consecutive years of risk losses. In each trial, a sequence of 50 loss samples is selected from the loss distribution, resulting in one hypothetical set of occurrences, or random walk, for the 50-year period. This process is repeated 10,000 times to generate the 10,000 random walks of 50 years duration each for the analysis. The sampling is done in such a manner that each year has a unique and statistically

9 Put another way, a desired level of solvency is a desired low probability or ruin, i.e., probability of a catastrophic loss exhausting an insurance companys capital.

2003 by CRC Press LLC

32-20

Earthquake Engineering Handbook

independent set of damage points; however, for each of the 50 years, all the 10,000 damage points are equally likely. The output of this model provides a probabilistic estimate of future claims, premiums, revenues, and prots (or losses) for each business strategy option that an insurance company may wish to consider.

32.5 Government Earthquake Insurance Pools


The value of catastrophe insurance in the recovery process following major disasters has long been recognized, and in some countries this has led to the development of government-controlled natural disaster insurance schemes. This section rst summarizes an overview of 15 government-controlled natural disaster insurance schemes listed in Table 32.5 [Scawthorn, 2001], not all of which protect against earthquake. Rather, a number of these schemes protect against ood, wind, or other natural hazards; in many cases, against all natural hazards. Nevertheless, many similarities exist irrespective of the hazard addressed, and much can be gained by reviewing their similarities and differences. The main trends in the programs surveyed may be summarized as follows: 1. Age. All of the programs dated from the 1960s or later, although the New Zealand Earthquake Commissions predecessor dated from the 1940s. 2. National vs. regional. About half of the programs were national in scope, the other half regional. Only the United States, France, and New Zealand appear to have truly national catastrophe programs (United States for ood only, New Zealand for earthquake and volcano, and France for multiperil). 3. Perils. Eleven of the programs covered ood, and six were multiperil). 4. Residential vs. commercial-industrial. All programs covered residential risks, 11 also covered commercial risks. In many cases, the commercial participation is minor. 5. Buildings. Coverage always included buildings and usually contents, only about one third covered time element (i.e., additional living expenses [ALE] or BI). Two programs covered publicly owned buildings. 6. Voluntary vs. compulsory. More of the programs were voluntary, with only three being compulsory. Compliance rates for the compulsory programs, while not available, did not appear to be near 100%. About half the programs premiums were risk based, and none was subsidized. i.e., all programs appeared to attempt to operate on a basis such that premiums covered claims and expenses, on average.

TABLE 32.5 Natural Hazards Insurance Programs Surveyed


No. 1 2 3 4 5 6 7 9 10 11 12 13 14 15 Country U.S. U.S. U.S. U.S. New Zealand France Spain Japan UK Czech Republic Switzerland Switzerland Switzerland Hungary Program Name Florida Hurricane Cat Fund Hawaii Hurricane Relief Fund California Earthquake Authority National Flood Insurance Plan Natural Disaster Fund Cat Nat, Caisse Centrale de Reassurance Consorcio Compensacin de Seguros Japan Earthquake Reinsurance Company

Interkantonaler Rckversicherungsverband Elementarschadenpool Schweizer pool fr Erdbendeckung

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-21

7. Mitigation. Five of the programs offered incentives or positively considered mitigation. However, only two of the programs actively used hazard maps in their underwriting or loss control (one of these, the U.S. National Flood Insurance Program (NFIP), has a very active hazard mapping program). 8. Deductible. Most programs had a deductible, which was typically 10% or less of TIV. Limits on coverage varied between a percentage of TIV, or a at limit, but the limit was 100% replacement cost in about half the programs. 9. Management. About one third (six) of the programs were publicly (i.e., government) managed, while about two thirds (ten) were privately managed. It should be noted that in some cases the private management was by a nonprot quasipublic board, independent of the government. Distribution and servicing of insurance policies were typically through the established distribution networks of private primary insurance companies and agents. 10. Primary vs. reinsurance: Four of the programs were insurance (i.e., primary) pools, and four were reinsurance pools. Virtually all programs covered second and subsequent losses. Only two programs were directly funded by the government. While there are a number of government pools in existence for natural hazards, there are four that dominate earthquake risk for government pools: the California Earthquake Authority, the Japan Earthquake Reinsurance Company, the Turkish Catastrophic Insurance Pool, and New Zealands Earthquake Commission. In large part, this is due to the combination of market size and earthquake hazard. The United States accounted for 32.35% of the world s total insurance premium volume in 1997, with nonlife premiums at U.S. $897 billion accounting for 42% of total premiums. The great preponderance of earthquake risk in the United States is in California. Japans share of world markets was 23.05%. In this section, we describe these programs; additional information is available in Guy Carpenter [2001].

32.5.1 California and the California Earthquake Authority


In 1996, the California legislature established the CEA (California Earthquake Authority) as a privately nanced, publicly managed entity to help California residents protect themselves against earthquake loss. Today, the CEA is the worlds largest residential earthquake insurer. Acting through its participating insurers, the CEA sells earthquake policies to homeowners, mobile-home owners, condominium owners, and renters throughout California, and provides retrot assistance to help people protect their homes against earthquakes. This section discusses earthquake insurance in California, from early in the twentieth century to the creation of the CEA. 32.5.1.1 California: Early Earthquake Insurance Experience California is well known, especially to insurers, for having earthquakes. The 1906 San Francisco earthquake caused approximately $300 million in losses, about 80% of which was due to the re that followed, and virtually all of the re-related losses were insured. This was the largest insurance catastrophe up to that time, and remained so for many years. Many companies struggled to pay off their claims, in some cases taking years. Yet damage-causing earthquakes have not been common in California until recently. After 1906, the next damage-causing earthquakes occurred in Santa Barbara (1925) and in Long Beach (1933) but were not especially catastrophic for insurers. The modern era of earthquake activity began with the 1971 San Fernando earthquake, which caused substantial damage to homes, businesses, and public buildings. Insurance covered only a small portion of the losses because very few people had earthquake coverage, and there was very little re damage. (In California, as well as other states, homeowners insurance covers any re losses, even if caused by an earthquake.) The 1971 disaster caused several important changes in attitudes. It raised the consciousness of people with respect to the dangers of earthquakes, and it focused attention on the necessity of revising the building codes to require earthquake-resistive design and construction. The event also reinforced the insurance industrys concern about the threat of earthquakes, and prompted the California Insurance

2003 by CRC Press LLC

32-22

Earthquake Engineering Handbook

Department to request private insurers to report annually on their earthquake exposures. This questionnaire, still in use today, is valuable to the department in monitoring the solvency of insurers. 32.5.1.2 The Northridge Earthquake Although a series of damaging earthquakes have occurred in California since the San Fernando earthquake, none compares to the January 17, 1994 Northridge earthquake. Next to Hurricane Andrew, it is the largest insured natural disaster ever, causing the most insured damage by far of any earthquake in the United States, with total insured losses of more than $12.5 billion. (In contrast, only a fraction of the damage caused by the Kobe, Japan earthquake in 1995 was covered by insurance.) To put the $12.5 billion gure in context, the population of Los Angeles County (where Northridge is located) is 9,244,646, with 1,565,862 single-unit dwellings and 1,402,997 apartment units (as of January 1, 1995). This works out to insurance company payments of $1352 for each man, woman, and child over a very large geographic area. The value of the total loss from the earthquake, including disaster assistance and uncompensated losses, is much greater. The insured loss amounts were about one third commercial and two thirds residential. The high peak ground acceleration of the earthquake was the primary cause of damage to building contents, chimneys, and garden walls. Building damage from landslide and liquefaction was not as common as in the Loma Prieta earthquake. While the insurance payments on this event exceed all of the earthquake insurance premiums collected in the twentieth century, this amount should only be compared to the current premium volume because so few structures were insured in past years, and so little premium collected. A typical earthquake policy insures for loss against structural damage, damage to contents, and loss of use (residential) or business income (commercial). Loss of use covers the cost of hotel accommodations and meals while the structure is being repaired, or it covers the loss of rental income on the house. Business income covers the loss of prots and the costs arising from shutting down the business (sometimes called business interruption). In the Northridge earthquake, for every $100 of insured residential damage, there was an average of $20 of content damage, and $10 of loss of use. It turned out that these ratios were the same for the 1989 Loma Prieta earthquake, even though the dollar amounts were much greater in Northridge. The insurance industry had great difculty estimating the ultimate losses from the Northridge earthquake. Property Claim Services (PCS) collects statistics on disasters and runs training classes on catastrophe claims handling for member insurance companies. PCS was the designated insurance-industry collector of the estimates of the Northridge insured losses, as well as the media spokesperson for this catastrophe. It polled its members and announced estimates of the total industry-insured losses periodically. Table 32.6, which provides a summary of these gures, shows that estimates of total insured losses grow over time as more accurate damage gures become available.

TABLE 32.6 Estimates of Insured Losses Caused by the Northridgc Earthquake, January 17, 1994
Month of Estimation vs. Estimate of Total Losses (in billions of dollars) 1994 January April June August October 1995 January March May Source: Property Claim Services.
2003 by CRC Press LLC

2.5 4.5 5.5 7.2 9.0

10.4 11.2 11.7

Insurance and Financial Risk Transfer

32-23

The California Insurance Department conducted its own surveys. The rst survey asked insurers to report their estimates as of OctoberNovember 1994, and the total reported was $8.8 billion. The survey conducted FebruaryMarch 1995 reported that total losses were $10.6 billion, including loss adjustment expenses. Both of these surveys were consistent with the survey data shown in Table 32.6. A short, nal survey made in 1996 conrmed that the nal gure was $12.5 billion for industry-insured losses. Earthquake damage losses are difcult to estimate because the full extent of damage is not known until reconstruction and repair have been completed. This is why the initial estimates after the earthquake were about $2.5 billion and then grew monthly for more than 2 years [Scawthorn, 1995]. 32.5.1.3 The California Earthquake Authority (1996) In 1985, the California legislature passed a law requiring insurers writing homeowners insurance on oneto four-family units to offer earthquake coverage on these structures. There was no requirement that the owners had to buy earthquake insurance, only that the insurers had to offer it. This law was not instigated by consumers but was passed at the urging of the insurance industry to overcome a recent lower court decision that had required greatly expanded coverage under the homeowners policy, including earthquake damage. In other words, the lower court decision appeared to enable homeowners to collect for earthquake damage even though they had not purchased an earthquake endorsement, and the insurer had not received a premium for the earthquake coverage. The California Supreme Court eventually overturned the lower court case but the mandatory offer law remained in effect. Many people who were close to the drafting of this law believe that it was a poor solution to a legal problem that no longer existed. Furthermore, it took away from insurers the ability to manage their total earthquake exposure and forced them to insure structures that are so old or in such poor condition that they should not be provided with coverage. The insurance companies lived with this law until the 1994 Northridge earthquake struck. The insured damage was beyond expectations, and when homeowners across the state heard about the average insurance payments of $30,000 to $50,000 after the 10% deductible for homes in the Northridge area, earthquake insurance became a desirable commodity in the minds of many. The insurance companies reevaluated their earthquake exposures up and down the state and decided that they could not risk selling any more earthquake policies. In view of the mandatory offer law, the only legal response they had was to stop offering new homeowners policies, although in fact they did renew existing homeowners policies even at the risk of the policyholders opting to buy earthquake insurance. In view of the desire of homeowners to maintain their earthquake coverage in the future, there was no possibility that the mandatory offer law could be repealed by the legislature. The California Insurance Department surveyed insurers and found out that up to 90% of them had either stopped selling new homeowners policies or had placed restrictions on selling them. After extended discussions between the California Insurance Department and the large insurers, an advisory group of insurers and actuaries proposed the formation of a state-run earthquake insurance company, the CEA. The challenge from the start was to capitalize the CEA adequately. This challenge generated some innovative nancing ideas. First, the department and the advisory group listed the possible sources of funding and chose these possibilities: (1) cash up front, (2) post-event assessments on insurers, (3) post-event assessments on CEA earthquake policyholders, (4) reinsurance, and (5) borrowing in the capital market. Next, each of these possibilities was designated as a block, then the blocks were put into various patterns to decide which one made the most sense. In 1995, the state legislature passed a law, referred to as Assembly Bill 13, authorizing the insurance commissioner to undertake a feasibility study of the CEA proposal. The nal block pattern, which was incorporated in this law, is shown in Table 32.7, and totals $10.5 billion in start-up funding. The innovative features of this nancing plan are the ability to pay for a large earthquake while committing relatively few dollars up front. At the time of the formation of the CEA, the capital market had never been used to back potential earthquake losses. In the proposed plan, capital markets were given the opportunity to cover the layer of losses from $7 to $8.5 billion. Before the markets had a chance to raise the full amount of capital, Warren Buffet of Berkshire Hathaway offered to cover the layer with his companys funds, and the insurance commissioner accepted his offer. The excess-of-loss reinsurance
2003 by CRC Press LLC

32-24

Earthquake Engineering Handbook

TABLE 32.7 Capacity Participation in the CEA


Layer Up to $1 billion $3 billion $2 billion $1 billion $1.5 billion $2 billion Source of Funding Industry contingent assessment (to start the program) Industry contingent assessment (after the earthquake) Reinsurance (no reinstatement) Policyholder contingent assessment Capital markets Industry contingent assessment (after the earthquake) Total (in $ billions) 1 4 6 7 8.5 10.5

commitment was $2 billion in excess of $4 billion. No reinsurance commitment this large has ever been made. A group of reinsurance brokers was chosen to determine if such a large placement was possible. Presentations were made to reinsurance companies in the United States and Europe. A long list of reinsurers agreed to commit small amounts, which added up to $1.7 billion. The reinsurance that was placed was a 2-year policy with an aggregate limit, a provision that there would be no coverage after the limit had been exceeded, and a provision that the rate would be just under 15% rate on line (that is, 15% of the $1.7 billion, or $255 million, per year). This translates to very expensive reinsurance but was the only choice, because the worldwide reinsurers were already heavily committed to reinsuring the commercial earthquake insurance market in California. The all-residential CEA program would have to be in addition to the reinsurers commitment to the commercial earthquake insurance market in California. The $10.5 billion total funding amount assumed that 100% of the current earthquake insurance policyholders would agree to buy coverage insured through the CEA, which would be limited coverage policies with a higher deductible (15% instead of 10%) and additional exclusions. If the CEA had been in business at the time of the Northridge earthquake, the CEA would have had to pay out about $4 billion. Therefore, the $10.5 billion gure was chosen so that the CEA would be able to pay for an earthquake two-and-a-half times the Northridge event.

32.5.2 Japan and the Japan Earthquake Reinsurance Company


In Japan, earthquake insurance has only been available since the 1950s. Residential earthquake insurance tends to be regarded as poor value for money and has limited use. This was highlighted following the 1995 Kobe earthquake for which the estimated direct property losses were of the order of $100 billion, of which the insurance industry provided less than $5 billion.10 At the time of the Great Hanshin-Awaji earthquake, only 3% of the residents of Kobe had any form of earthquake insurance. This was due in part to the popular perception that Kobe was a low to moderate earthquake risk. Demand has increased since the 1995 earthquake and, as of 2000, it was estimated that 15% of residents in Japan have earthquake insurance. Residential earthquake insurance is only available as a special extension provided by the re insurer but reinsured through the Japan Earthquake Reinsurance Company (JER). However, all policies include an earthquake re expenses insurance (EFEI) extension, which pays an additional 5% of the total sum insured if more than 50% of the dwelling is destroyed by re following earthquake or volcanic eruption. JER was formed in 1966 and is partly owned by the domestic insurance industry and partly by the government. JERs sole function is to reinsure the government-sponsored earthquake protection scheme for residential risks. For residential earthquake coverage, an insured may elect to purchase a separate earthquake policy issued by the policyholders household insurer covering earthquake shock and re, volcanic eruption, and tidal wave following earthquake. Under the latest revision of the scheme (January 1, 1996), the

10 Contrast this with Northridge, where total losses were estimated to have been $30 to $40 billion, and insurance payments were approximately $15 billion, or almost 50%.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-25

TABLE 32.8 Japanese Residential Earthquake Policy Terms


Extent of Damage Buildings Damage > 50% of sum insured (or 70% of oor area) 20% < damage < 50% of sum insured (or 20% < oor area < 70%) 3% < damage < 20% Contents 80% sum insured < damage 30% < damage < 80% of sum insured 10% < damage < 30% Paid (%) 100 50 5 100 50 5

maximum insurance limits are 50 million on buildings and 10 million on contents. Within these monetary limits, Table 32.8 indicates the percentages of the actual sum insured that the policy pays. All household policies and most commercial re policies, except warehouse risks, are automatically extended to include earthquake re expenses insurance without additional premium. The EFEI extension pays an additional 5% of the sum insured in the event that 50% or more of the property is destroyed by re following earthquake or volcanic eruption. The maximum EFEI indemnity is limited to 3 million with respect to domestic risks and 20 million with respect to industrial risks. Residential earthquake risks underwritten by insurance companies are reinsured exclusively with JER and shared with the Japanese government. Under the current arrangement, all residential earthquake business is reinsured 100% with JER, which retrocedes a certain portion of the portfolio back to the direct market. The remainder of the portfolio is guaranteed by the government. The aggregate limit of indemnity payable by all insurers and the government was increased to 4100 billion for any one occurrence with effect from March 31, 1999. Claims are scaled down pro rata if the insured losses from any one earthquake exceed the scheme limit.

32.5.3 Turkey and the Turkish Catastrophic Insurance Pool


The Turkish Catastrophic Insurance Pool (TCIP) was established by the Turkish government in cooperation with the World Bank, as a result of the 1999 Marmara earthquake. TCIP objectives are to: 1. 2. 3. 4. 5. Ensure most domestic dwellings have earthquake insurance Reduce government scal exposure Transfer most of catastrophe risk to international reinsurance and capital markets Build up TCIPs capital base over time to insure against larger events Encourage risk mitigation and safer construction practices

In order to accomplish this, in summary, the TCIP program consists of: 1. Compulsory earthquake coverage for all registered residential dwellings and independent business ofces in those same buildings 2. A stand-alone product, separate from re (homeowners) insurance but covering re, explosion, and landslide following earthquake 3. Coverage up to TL 20 billion (approximately $16,000 at year 2002 exchange rates) per dwelling (none for contents). Limits and sums insured valuations are expected to be adjusted annually based on TL ination rates. Losses are to be settled on the basis of current replacement cost (in TL) at the time of loss, subject to policy limit 4. Coverage in excess of TCIP, obtainable from private insurers 5. Policies distributed by private insurers acting as agents, i.e., assume no risk of loss 6. Elimination of penny claims and reduced administrative and reinsurance costs of the pool, a deductible of 2% introduced 7. Online (Web-automated) policy underwriting, in addition to insurers own underwriting systems, employed together with a centralized, integrated data management
2003 by CRC Press LLC

32-26

Earthquake Engineering Handbook

8. Functions that are completely outsourced to the private sector 9. Independent (hired by TCIP) loss adjusters, used in claim settlement 10. Fifteen rating categories, based on hazard zone (not CRESTA zones) and the type of buildings, as shown in Table 32.9.
TABLE 32.9 Turkish Catastrophic Insurance Pool Earthquake Insurance Rates
Zone (%) Type of Construction Steel and reinforced concrete buildings Masonry buildings made of stone or brick Other buildings I 2.00 3.50 5.00 II 1.40 2.50 3.20 III 0.75 1.30 1.60 IV 0.50 0.50 0.70 V 0.40 0.40 0.50

TCIP policy count and exposures were projected for October 2002 to be 2.5 million and total sums insured of $32 million, making it a rival of the California Earthquake Authority to be the largest earthquake insurance company in the world.

32.5.4 New Zealand and the Earthquake Commission


In 1941, the New Zealand government set up a special war damages insurance pool funded by a levy of 0.25% of insured value on all re insurance premiums. The government soon extended the scheme to cover earthquake, establishing the pool as the Earthquake and War Damage Fund. After the war, the levy was reduced to 0.05%. The scheme covered all property for indemnity value and, over time, was extended to other uninsured natural hazards. In 1993, it was restructured as the Earthquake Commission (EQC), and currently offers replacement coverage home insurance up to NZ $100,000 for buildings and $20,000 for contents, at a levy of 0.05% on all re insurance premiums.

32.6 Alternative Risk Transfer


This section discusses the use of earthquake modeling in a typical insurance-linked securitization, or alternative risk transfer (ART). In the second half of the 1990s, a major innovative alternative to traditional reinsurance emerged, which may be generically termed catastrophe securitization. A cat bond requires the investor to pay money up front that will be used by the rm if some type of triggering event occurs.11 In exchange for a high return on investment, the investor faces the possibility of losing either interest on the money in trust, or a portion or entire principal investment. The amount paid out to the rm (i.e., the ceding company) depends on how the cat bond is constructed, and this amount is specied in advance of the triggering event. Notably, the rm does not face any credit risk from the cat bond; the money to pay for the losses is already in hand. Initial insurance-linked securitizations were structured to mimic excess-of-loss reinsurance coverages. Thus, incurred monetary losses above an attachment point were paid out by the bondholders, i.e., the payment was indemnity-based. However, due to investor concerns related to moral hazard and data quality issues, the ceding company was required to retain a portion of the risk. So in some circumstances, the PCS industry loss became a more attractive loss trigger, especially if the ceding company did not have detailed policy information, or did not want to publicly disclose its book of business to competitors.

The investors up-front investment is generally placed in a trust, and paid to the ceding company in the case of a triggering event. This implies that the ceding company is only able to reinvest proceeds from cat bond premiums in liquid securities at approximately the risk-free rate.
2003 by CRC Press LLC

11

Insurance and Financial Risk Transfer

32-27

32.6.1 Index-Based Cat Bonds


Most of the cat bonds issued today are tied to a disaster-severity index (e.g., covering damage from a certain earthquake magnitude event within a specied region) rather than to the rms actual losses.12 Because these indexed parameters are normally independent of the rms actual losses, payments can be made to the rm immediately after the disaster occurs rather than being subject to the time delay necessary to compute actual losses, as in the case of insurance or reinsurance. Hence, indexed cat bonds reduce the amount of moral hazard in loss estimation. On the other hand, such a cat bond creates basis risk. Basis risk refers to the imperfect correlation between the actual losses suffered by the rm and the payments received from the cat bond. Insurance sold to rms or excess-ofloss reinsurance to insurers has very little basis risk because there is a direct relationship between the loss and the payment delivered by the reinsurance instrument. In May 1999, a parametric-based contract to cover the loss from an earthquake was purchased by Oriental Land, a Japanese company that is best known as the owner and operator of Tokyo Disneyland. This cat bond provides $100 million to the company if an earthquake of a specic magnitude occurs in the vicinity of Tokyo. The Japanese Meteorological Agency determines the measurement of event magnitude. The magnitude that qualies a given quake for payments to Oriental Land is higher as the location of its epicenter becomes more remote from Tokyo Disneyland. These bonds represent the rst direct access of the capital markets by a corporation seeking catastrophe risk nancing [Standard and Poors, 2000].

32.6.2 Other Types of Cat Bonds


Further variations on nonmonetary loss triggers have also been introduced, including loss triggers based on modeled losses. When an event takes place, the loss to investors is calculated by inputting event parameters into the same model used to perform the risk estimate. (In order to protect the investor, the model is placed in escrow for the duration of the term of the security.) The model then calculates the payout to the issuer. A further variant of nonmonetary triggers has been recently introduced, called second generation parametric triggers, in which the loss trigger is dened by a specic physical parameter, such as peak ground velocity, and measured (or calculated) at a dened series of locations in the vicinity of the insured exposures. In many cases, the physical parameters are available on the Internet within minutes of the event, permitting investors and issuers to quickly determine if a payout is forthcoming. In summary, nonmonetary triggers rely on portions of the loss models to estimate the loss exceedance probabilities, thus creating a simple structure for investors to understand. This also reduces issues of moral hazard and uncertainty in damage and loss payments. Of course, the issuer then retains those risks in the form of basis risk. In conclusion, probabilistic loss models have become essential ingredients in ART earthquake insurance transactions, either for monetary or nonmonetary loss triggers.

32.6.3 Cat Bond Example


A California earthquake transaction is used as an example. California represents one of the largest sources of earthquake risk in a reinsurers portfolio, and hence alternate risk market sources are appealing. Since 1997, ve California earthquake securitizations totaling over $700 million have been placed. In addition, several more multiperil and multiregion deals that included California earthquake have also been completed. A typical transaction involves a reinsurer with excess California earthquake exposures. Such a company wishes to hedge some of that exposure to large events, such as a repeat of the 1906 California earthquake. Because their reinsurance portfolio is likely to derive from a large number of primary insurance com-

12

For more details on the structure of recent cat bonds, see Insurance Services Ofce [1999].

2003 by CRC Press LLC

32-28

Earthquake Engineering Handbook

0.8

Aggregate Probability (%)

Attachment Amount

0.7 0.6

0.5 0.4 Exhaustion Amount

0.3 21.5

23.5

25.5

27.5

29.5

31.5

Cumulative Insured Losses ($ in billions)

FIGURE 32.10 The aggregate loss exceedance curve for California-insured exposures.

panies with California earthquake exposures, their losses are expected to be highly correlated with reported industry losses. So instead of using their own portfolio, they select an index based on Property Claims Services,13 which regularly reports insured U.S. property losses from catastrophic events. In such a transaction, the ceding company retains the services of a risk-modeling consultant to perform the analysis on the portfolio, in this case earthquake-insured exposures in the State of California. The modeling company compiles an estimate of insurance in force, using data from the California Department of Insurance, California Earthquake Authority, private insurers, tax assessors, and census data. A portfolio of insured properties by zip code is created and input into the loss model. The portfolio denes the type of occupancy, location (zip code), and underlying insurance policy coverages such as deductibles, limits, sublimits, etc. The portfolio is analyzed using a probabilistic earthquake model, and the loss exceedance curve is output from the model. Based on the reinsurers desired level of protection and correlation with the industry loss, the reinsurer selects a level of coverage desired. from the curve. The resulting attachment and exhaustion probabilities are determined from the curve, and the expected loss for the layer is derived by integrating the area under the curve between those two points. Figure 32.10 shows the curve from a typical transaction. The expected loss to the layer becomes the base estimate for pricing the security. Moodys Investor Service [1997] and other rating agencies have developed criteria to rate insurance-linked notes based on comparing model-based loss probabilities with historical credit default and loss. Following a series of stress tests by the rating agencies, the agency determines a bond rating, which then becomes the basis for pricing based on spreads over LIBOR (London Interbank Overnight Rate) for equivalent rated securities.

32.6.4 ART Summary


By transforming cat risk into a security and trading it in the global capital markets (Wall Street, London, Tokyo), the total capital that can be accessed is one to two orders of magnitude larger (i.e., U.S. $5 to $50 trillion) [Froot et al., n.d.]. Additionally, cat bonds are relatively uncorrelated with normal stocks and bonds, which makes them attractive to investors. For these reasons, cat bonds have emerged as an increasingly attractive method for reinsuring natural hazards cat risks, such as ood, windstorm, and earthquake. Traditional reinsurers have recognized this, and are ready to assist their clients in the complexities of issuing a cat bond. One of the major complexities is the increased public scrutiny associated

13

Property Claims Services, a division of ISO Services, www.iso.com/AISG/pcs.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-29

with a publicly traded security. That is, in traditional reinsurance, the contract is of a traditional nature, well understood by both parties, and not normally subject to public scrutiny. In the case of cat bonds, security regulations in the U.S. and other capital markets require substantial disclosure, and investors require attractive rating by agencies such as Moodys, Standard and Poors, and others, which closely review the details of the cat bond and its underlying analyses. Another form of ART, termed a swap, is available also for nontraditional nancial risk transfer. A swap is a contract between two parties, specifying payments under certain conditions. Used for several decades for hedging foreign currency or interest risk, swaps are also now being used to hedge cat risk. Currently most cat risk swaps are private transactions, usually only between insurers.

32.7 Summary
The insurance industry is a large, complex endeavor that, for several hundred years, has provided nancial protection against common risks such as re and theft. Insurance has also been an effective tool for loss prevention, via both the market as well as the institution of systematic loss prevention programs by companies such as the Factory Mutual system. Due to a clear understanding of earthquake causes and effects only emerging in the twentieth century, earthquake insurance for much of the century had a much softer foundation than more common forms of insurance, and the question was often asked, Are earthquakes even an insurable risk? In this context, Freemans little-known, pioneering work, Earthquake Damage and Earthquake Insurance [Freeman, 1932] deserves to be mentioned. In that treatise, Freeman not only collated existing knowledge on earthquakes but also analyzed and provided the basis for earthquake insurance rates for California and other locations, using sound underwriting principles. Today, earthquakes and their effects are better understood, and earthquake insurance is increasingly better founded, available and rationally priced using risk-based underwriting based on earthquake loss modeling. Finally, new techniques such as ART are also emerging, which should increasingly lower the cost of earthquake insurance.

Dening Terms
Adverse selection A situation in which an insured or group of insureds obtain insurance contracts
with adverse facts being unknown to the insurer. In earthquake insurance, persons with wellconstructed houses know their houses to be low risk, and therefore are less likely to purchase earthquake insurance, while persons with houses likely to sustain a large loss are more likely to purchase insurance. If the insurer does not diligently determine the specic earthquake-related quality of the houses insured, the resulting portfolio will have a lower quality (higher loss) than that on which the insurance premiums are based, and adverse selection has occurred. ALE Additional living expenses. Payments made to insureds for expenses associated with loss of use of their normal residence. Alternative risk transfer New nancial mechanism for transferring risk, employed as an alternative to traditional insurance. Examples include swaps and cat bond securitizations. Amortization period Synonymous with payback period, this term is used in the rating of per occurrence excess coverage and represents the number of years at a given premium level necessary to accumulate total premiums equal to the indemnity. Assume To accept all or part of a ceding companys insurance or reinsurance on a risk or exposure. Attachment point The amount at which excess reinsurance protection becomes operative; the retention under an excess reinsurance contract. Base premium The ceding companys premiums (written or earned) to which the reinsurance premium rate is applied to produce the reinsurance premium. Basis risk The risk associated with the difference between the best estimate of a parameter (e.g., loss in an earthquake) and the actual value of the parameter.

2003 by CRC Press LLC

32-30

Earthquake Engineering Handbook

BI Business interruption. Payments made to commercial entities for expenses associated with restoring normal operations.

Burning cost The ratio of actual past reinsured losses to ceding companys subject matter premium
(written or earned) for the same period. Used to analyze past reinsurance experience or to project the future. The matching of all losses incurred within a given 12-month period, usually beginning on January 1, with all premium earned within the same period of time. Capacity A company that is wholly owned by another organization (generally noninsurance), the main purpose of which is to insure the risks of the parent organization. Catastrophe A form of excess-of-loss reinsurance that, subject to a specic limit, indemnies the ceding company against the amount of loss in excess of a specied retention. This loss amount represents the accumulation of losses resulting from a catastrophic event or series of events. To transfer to a reinsurer all or part of the insurance or reinsurance written by a ceding company with the object of reducing the potential liability of the latter. Catastrophe reinsurance In reinsurance, an allowance (usually a percentage of the reinsurance premium) made by the reinsurer for part or all of a ceding companys acquisition and other costs. The ceding commission may also include a prot factor. Cede A written statement issued by an intermediary, broker, or direct writer, indicating that coverage has been effected. Cession The unit of insurance passed to a reinsurer by the primary company that issued a policy to the original insured. A cession accordingly may be the whole or a portion of single risks, dened policies, or dened divisions of business, as agreed in the reinsurance contract. Commercial lines Types of insurance offered to larger organizations, such as companies, corporations, municipalities, etc., for protection against re, theft, business interruption, loss of valuable papers, and other risks. DFA Dynamic nancial analysis. An analysis based on stochastic simulation of nancial operations over time. In the insurance context, the premium income, claims paid, investment income, and other cash ows are stochastically simulated over a time horizon to determine risk adjusted return on capital (RAROC), probability of ruin, or other parameters of interest. Earned premium The portion of a premium which is the property of an insurance company, based on the expired portion of the policy period. Excess-of-loss reinsurance A generic term describing reinsurance that, subject to a specied limit, indemnies the ceding company against all or a portion of the amount in excess of a specied retention. Exposure The state of being subject to the possibility of loss; the extent of risk as measured by various standards such as payroll, gate receipts, and area. Facultative The reinsurance of part or all of (the insurance provided by) a single policy in which each cession is negotiated separately. The primary company and the reinsurer have the option of accepting or declining each individual submission (as parties agree in treaty reinsurance). Frequency Number of times an event such as a loss occurs. GIS Geographic information system. Computer technology permitting relational database analysis of data with a geographic attribute. Commonly used to identify spatial trends, which are presented in mapped or statistical forms. Ground-up loss The total amount of loss sustained before deductions are applied for reinsurance coverage that inure to the benet of the cover being considered before the application of a deductible, if any, because that base theoretically reects changes in exposure. Guaranty fund A fund supported by assessments against solvent insurance companies to absorb losses of claimants against insolvent insurers. Incurred losses In insurance accounting, an amount representing the losses paid plus the change (positive or negative) in outstanding loss reserves within a given period of time.

2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-31

Indemnity The compensation owed by the insurer to the insured, per the terms of the insurance
contract. Indemnity makes the insured whole but nothing more.

Insurance Mechanism for contractually shifting burdens of a number of pure risks by pooling them. Law of large numbers A mathematical concept that postulates that the more times an event is
repeated, the more predictable the outcome becomes.

Leveraged effect The disproportionate result produced by ination on a reinsurers liability in excess
of loss reinsurance compared with the ceding companys liability. Inationary increase in average claim costs of a reinsured usually produced even greater for its excess of loss reinsurer, because an increase affecting all losses multiplies itself when affecting the excess of loss portion above that retention limit. Liability insurance Covers damage on behalf of an insured who becomes legally obligated to pay because of an actual or alleged negligent act and omissions. LIBOR London Interbank Overnight Rate. A standard interest rate used in the nancial industry. Line Either the limit of insurance to be written which a company has xed for itself on a class of risk (line limit), or the actual amount that it has accepted on single risk or other unit. Lines of business Five primary sectors of insurance: life, health, annuity, property, and liability. Loss exceedance curve A graph depicting the probability of exceeding a specied level of loss. Loss ratio Losses incurred expressed as a percentage of earned premiums. Loss reserve For an individual loss, an estimate of the amount the insurer expects to pay for the reported claim. For total losses, estimates of expected payments for reported and unreported claims. May include amounts of loss adjustment expenses. Loss reserve transfer A type of reinsurance treaty through which companies can exchange certain types of liabilities. Also known as loss portfolio transfer. MFO Most favored option. The option that appears most satisfactory, considering all aspects. Net loss The amount of loss sustained by an insurer after making deductions for all recoveries, salvage, and all claims upon reinsurers with specics of the denition derived from the reinsurance agreement. Such net loss may or may not include claim expenses. As provided in the reinsurance agreement, net loss can be conned to the amount paid by the reinsured within applicable policy damages in excess of applicable policy limits because of failure of the reinsured to settle within applicable policy limits. Net retention The amount of insurance that an insurer keeps for its own account and does not reinsure in any way. Nonproportional reinsurance Reinsurance under which the reinsurers participation in a loss depends on the size of the loss. Also known as excess-of-loss reinsurance. Occurrence An event that results in an insured loss. In some lines of insurance, such as liability, it is distinguished from an accident in that the loss does not have to be sudden and fortuitous. It can result from continuous or repeated exposure. Parametric trigger Catastrophe securitization in which payments are triggered by the occurrence of a parameter (e.g., of a specied magnitude event within a specied region), and not the ensuing monetary loss. Penetration Also known as the buy rate, market penetration is dened as the percentage of insured vs. the maximum number of possible insured. Personal lines Types of insurance (homeowners, auto) sold to individuals to protect their personal property. Pool Any joint underwriting operation of insurance or reinsurance in which the participants assume a predetermined and xed interest in all business written. Pools are often independently managed by professionals with expertise in the classes of business undertaken, and the members share equally in the premiums, losses, expenses, and prots. An association and a syndicate (excluding that of Lloyds of London) are both synonymous with a pool, and the basic principles of operation are much the same.
2003 by CRC Press LLC

32-32

Earthquake Engineering Handbook

Portfolio A dened body of insurance policies in force. Premium The monetary consideration in contracts of insurance and reinsurance. Primary An adjective applied in reinsurance to these nouns: insurer, insured, policy, and insurance,
referring to the parties involved in the initial insurance transaction with the property owner.

pdf Probability density function. PML Variously dened but usually intended to indicate an estimate of maximum loss which can be
sustained under realistic situations.

Pure premium That part of the premium that is sufcient to pay losses and loss adjustment expenses
but not including other expenses.

Pure risk Situation involving a chance of a loss or no loss but no chance of gain. For example, either
ones home burns, or it does not; this risk is insurable. Pure risk is also synonymous with pure premium. RAROC Risk adjusted return on capital. The present value of capital, taking into account the stochastic character of risks to that capital, commonly determined via a dynamic nancial analysis. Rate The percentage or factor applied to the ceding companys base premium to produce the reinsurance premium, or the percentage applied to the reinsurers premium to produce the commission payable to the primary company (or, if applicable, the reinsurance intermediary). Reinsurance The transaction whereby the reinsurer, for a consideration, agrees to indemnify the ceding company against all or part of the loss that the latter may sustain under the policy or policies it has issued. Reinsurer An organization that assumes the liability of another by way of reinsurance. Retention The amount that an insurer assumes for its own account. In pro rata contracts, the retention may be a percentage of the policy limit. In excess-of-loss contracts, the retention is a dollar amount of loss. Retrocession The transaction whereby a reinsurer cedes to another reinsurer all or part of the reinsurance it has previously assumed. Risk The chance of loss. Also used to refer to the insured or to property covered by a policy. In reinsurance, each company makes its own rules for dening units of hazard or single risks. Ruin Lack of solvency (i.e., bankruptcy), where liabilities exceed assets and ability to honor liabilities cannot be achieved. Securitization The creation of a security, such as a bond. The security is issued in the capital markets, as a promise to pay a series of payments against the loss of principal and interest to the bondholder if a specied indemnity-based loss or parametric trigger occurs. Severity Size of the loss used as a factor in calculating insurance premium. Surplus to policyholders The net worth of an insurer as reported in its annual statement. For a stock insurer, the sum of its unassigned surplus and capital. Surplus-share reinsurance A form of pro rata reinsurance indemnifying the ceding company against loss for the surplus liability ceded. Essentially, this can be viewed as a variable quota share contract wherein the reinsurers pro rata share of insurance on individual risks will increase as the amount of insurance increases in order that the primary company can limit its new exposure regardless of the amount of insurance written. Syndicate A group of insurers or underwriters who join to insure property that presents high values or high hazards. Also see Pool. Time element Refers to insurance coverage for additional living expense or business interruption. TIV Total insured value. Treaty A reinsurance agreement between the ceding company and the reinsurer, usually for one year or longer, which stipulates the technical particulars applicable to the reinsurance of some class or classes of business. Treaty reinsurance A standing agreement between reinsured and reinsurer for the cession and assumption of certain risks as dened in the treaty.
2003 by CRC Press LLC

Insurance and Financial Risk Transfer

32-33

Underwriting capacity The maximum amount of money an insurer or reinsurer is willing to risk in
a single loss event on a single risk or in a given period. The limit of capacity for an insurer or reinsurer may also be imposed by law or regulatory authority. Unearned premium reserve The sum of all the premiums representing the unexpired portions of the policies or contracts that the insurer or reinsurer has on its books as of a certain date. It is usually based on a formula of averages of issue dates and the length of term. Write To insure, to underwrite, or to accept an application for insurance.

References
BestWeek. 1996. Earthquakes: A Major Paradigm Shift for P/C Insurers, BestWeek: Property Casualty Supplement, P/C 118, March 25. Camerer, C. and Kunreuther, H. 1989. Decision Processes for Low Probability Events: Policy Implications, J. Policy Anal. Manage., 8, 565592. Carpenter, G. 2001. The World Catastrophe Reinsurance Market, published by Guy Carpenter, New York. Report available on the Internet at http://www.guycarpenter.com/publications/pubmn.html. CDI (California Department of Insurance). 2001. California Earthquake Liability Questionnaire, California Administrative Code Title 10, Chapter 5, Subchapter 3, Article 3, Section 2307 General Instructions (Revised 10/2001), California Department of Insurance. http://www.insurance.ca.gov/ RRD/RSU/Forms/Year2001/PML2001/pml2001.htm. EQC (Earthquake Commission). 1998. Annual Report 199798, Earthquake Commission, Wellington, New Zealand. Freeman, J.H. 1932. Earthquake Damage and Earthquake Insurance, McGraw-Hill, New York. Froot, K. et al. n.d. The Evolving Market for Catastrophic Event Risk, Guy Carpenter, New York, www.guycarp.com. Heubner, S.S. and Black, K. 1995. Property and Liability Insurance, Prentice-Hall, New York. Hirschberg, J.C., Gordon, P., and Petak, W. J. 1978. Natural Hazards: Socio-Economic Impact Assessment Model, NSF/PRA-7509998/5, J.H. Wiggins and Co., Redondo Beach, CA. Insurance Information Institute. 2002. Available on the Internet at http://www.nancialservicesfacts.org/ nancial/insurance/overview/natureandsize. Insurance Research Council and Insurance Institute of Property Loss Reduction. 1995. Coastal Exposure and Community Protection: Hurricane Andrews Legacy, IRC, Wheaton, IL and IIPLR, Boston. Insurance Services Ofce. 1999. Financing Catastrophe Risk: Capital Market Solutions, Insurance Services Ofce, New York. King, S. and Kiremidjian, A. 1997. Use of GIS for Earthquake Hazard and Loss Estimation, in Geographic Information Research: Bridging the Atlantic, Taylor & Francis, London. Kunreuther, H. 1996. Mitigating Disaster Losses through Insurance, J. Risk Uncertainty, 12, 171187. Kunreuther, H. and Roth, R.J., Eds. 1998. Paying the Price: The Status and Role of Insurance against Earthquakes in the U.S., Joseph Henry Press, Washington, D.C. Kunreuther, H. et al. 1978. Disaster Insurance Protection: Public Policy Lessons, John Wiley & Sons, New York. Kunreuther, H., Hogarth, R., Meszaros, J., and Spranca, M. 1995. Ambiguity and Underwriter Decision Processes, J. Economic Behav. Organ., 26, 337352. Lecomte, G. and Gahagan, K. 1998. Hurricane Insurance Protection in Florida, in Paying the Price: The Status and Role of Insurance Against Earthquakes in the U.S., Kunreuther, H. and Roth, R., Eds., Joseph Henry Press, Washington, D.C., chap. 5. Long, J.D. and Gregg, D.W. 1965. Property and Liability Insurance Handbook, Dow Jones-Irwin, Homewood, IL. Moodys Investor Service. 1997. Approach to the Rating of Earthquake-Linked Notes, September 1997, Moodys Investor Service, New York.

2003 by CRC Press LLC

32-34

Earthquake Engineering Handbook

Palm, R. 1995. Earthquake Insurance: A Longitudinal Study of California Homeowners, Westview Press, Boulder, CO. Petak, W.J. and Atkinsson, A.A. 1982. Natural Hazard Risk Assessment and Public Policy: Expecting the Unexpected, Springer-Verlag, New York. Roth, R.J. Jr. and Van, T.Q. 1998. California Earthquake Zoning and Probable Maximum Loss Evaluation Program: An Analysis of Potential Insured Earthquake Losses from Questionnaires Submitted by Property/Casualty Insurers in California, per California Administrative Code, Title 10, Chapter 5, Subchapter 3, Section 2307. California Department of Insurance, Los Angeles. Scawthorn, C. 1981. Urban Seismic Risk: Analysis and Mitigation, Ph.D. dissertation, Kyoto University, Kyoto, Japan. Scawthorn, C. 1995. Insurance Estimation: Performance in the Northridge Earthquake, Contingencies, the magazine of the American Institute of Actuaries, Washington, D.C. Scawthorn, C. 2001. National Programs for Natural Hazards Insurance, First Annual IIASA-DPRI Meeting, Integrated Disaster Risk Management: Reducing Socioeconomic Vulnerability, International Institute for Advanced Systems Analysis, Laxenburg, Austria. SEAOC (Structural Engineers Association of California), Seismology Committee. 1999. Recommended Lateral Force Requirements and Commentary, Sacramento, CA. Standard and Poors. 2000. Sector Report: Securitization, June. Steinbrugge, K.V. 1982. Earthquakes, Volanoes and Tsunamis: An Anatomy of Hazards, Skandia America Group, New York. Steinbrugge, K.V. and Roth, R.J., Jr. 1994. Dwelling and Mobile Home Monetary Losses Due to the 1989 Loma Prieta, California, Earthquake with an Emphasis on Loss Estimation, U.S. Geological Survey Bulletin 1939-B. Stone, J. 1973. A Theory of Capacity and the Insurance of Earthquake Risks, I and II, J. Risk Insur., 40, 231243 (Part I) and 40, 339355 (Part II). Walker, G.R. 2000. Earthquake Engineering and Insurance: Past, Present and Future, Aon Re Australia, available on the Internet at http://www.aon.com.au/reinsurance/knowledge/quake_engineering.asp Weinstein, N., Ed. 1987. Taking Care: Understanding and Encouraging Self-Protective Behavior, Cambridge University Press, New York. Wiggins, J.H., 1981. Earthquake Hazard and Risk Mitigation, J.H. Wiggins Company, Report No. 80-1371-1, Redondo Beach, CA.

Further Reading
The pioneering work by Freeman [1932] still offers much of value. Several standard references on insurance in general, such as Long and Gregg [1965] or Heubner and Black [1995], provide more detail on insurance contracts, underwriting, and related matters. Steinbrugge [1982] is a good overview of the California PML methodology and earthquake insurance prior to the innovations of probabilistic earthquake loss modeling. Paying the Price [Kunreuther and Roth, 1998] provides an excellent overview of the current status of natural hazards insurance.

2003 by CRC Press LLC

Das könnte Ihnen auch gefallen