The Economic Consequences of a Catastrophic Earthquake: Proceedings of a Forum (1992)

The purpose of this chapter is to raise questions about the differential impacts (or effects) that earthquake events may have. Certainly, earthquakes of different magnitudes and intensities will have differential effects, as will seismic events that take place in different types of geologic areas (as were described by Dr. Hamilton in his presentation). Besides these geophysical conditions, however, it must be noted that the types of social and policy contexts in place in the communities which are struck by a major earthquake will also have an effect on the types of impacts that are sustained. These differential impacts may be related to the stage of the city's life cycle (whether it is old and aging or new and robust), the types of building stock in existence, the extent to which production is localized or dispersed nationally, the extent to which a major market is disrupted, the extent to which earthquake insurance is available and affordable, and the extent to which mitigation efforts have been undertaken.

The first presenter in this chapter is Professor Anthony M. Yezer from George Washington University. Dr. Yezer has a doctoral degree in economics from MIT, with an emphasis on applied microeconomic theory. HIS presentation will focus on measuring the effects of catastrophic earthquakes in different regions of the country.

The second presentation will be made by Professor Howard Kunreuther and is a joint effort with his coauthor, Professor Neil Doherty. Dr. Kunreuther is a professor of decision sciences and director of the Wharton Risk and Decision Processes Center at the University of Pennsylvania. In the recent past, he was director of the Decision Risk and Management Science Program at NSF. Dr. Doherty is a professor of insurance in the Wharton School at the University of Pennsylvania. The work that forms the basis for this presentation focused on the role of insurance compensation, incentive mechanisms, and regulation as policy tools to reduce the impacts of disasters. Their presentation will focus on the role of loss-reduction measures.

Unfortunately, there has been relatively little research on the topic of economic effects of serious earthquakes by economists. Thus, the statements that can be made are based largely on extension of economic theory that has been developed to analyze effects of phenomena analogous to earthquakes. Such applications of theory also provide perspective on the types of questions that should be asked in trying to assess economic effects of earthquakes.

First, a review of the literature on economic consequences of disasters must be conducted. Second, the issue is placed in general context of economic theory by analyzing the relationship between natural hazards and economic development of a region. A hazard event, such as a serious earthquake, has a direct and immediate effect on the capital stock of a region and on the physical health of its residents. Then there is a long-run effect that follows the event as the expectations for future productivity of the region change. It is important to consider both the immediate and the long-run effects when attempting to characterize economic effects of hazards on a region.

Third, evidence on how uncertain natural-hazard events enter an economic model of development of a regional economy is considered. Special problems arise in connection with infrequent and very uncertain events such as serious earthquakes. It is then possible to conduct exercises in which the regional economy responds to disaster events. Finally, implications for additional research, particularly directed to earthquake hazards, are presented.

Before beginning, it is important to frame the question of the Forum on Earthquake Economic Issues in economic terms. What is meant by "economic consequences of a catastrophic earthquake"? What is the alternative to having a serious earthquake in 1990? Is it having a serious earthquake in 1991? Is it never having a serious earthquake in the history of the world? Is it having two serious earthquakes in 1991 to make up for the one that was missed in 1990? The manner in which the question is posed is crucial for discussion of the problem of earthquakes and the notion of economic effects. Obviously, delay of an earthquake by one year is relatively trivial compared with reduction in the total number of serious earthquakes. However, it is not clear that we have the option of lowering the total number of earthquakes, and failure to have an earthquake this year may simply mean that it has been delayed. Economic agents are assumed to treat earthquakes as random events that occur according to a Poisson process in which the probability of having another earthquake is independent of the number of previous earthquake events.³¹ The occurrence or nonoccurrence of earthquakes this year provides information that is used in updating forecasts of future earthquakes.

There is literature that relates the asset prices of housing in a given area to proximity to a natural hazard.^32,33 The standard finding is that houses farther from the hazard sell for higher prices, and these appear to reflect differences in insurance costs. Recently, contrary evidence was published that indicates that appraisers, lenders, and buyers appear to ignore earthquake hazards,³⁴ The announcement of a possible future disaster in Mammoth, California, had important effects on property values.³⁵ Such results suggest that there is a market response to disasters. These studies have little or no

dynamic component and have been done for single areas in which endangered land could be compared with safe locations.

A 1985 report to NSF provided a brief but cogent review of earlier evidence on the longer-term economic effects of disasters.³⁶ The tests provide very mixed evidence on economic effects of disasters.³⁷ The differences in the literature appear to match an extensive list of case studies on large disaster incidents against an econometric estimate of long-run effects on housing markets and survey evidence of local officials.^{38,39,40,41,42,43,44,45}

Case studies of large disaster events provide great detail and document the importance of individual area responses. They often find that the disaster only interrupts economic trends and is followed by a continuation of the predisaster economic decline or advance. In some cases, substantial changes in the growth path of the local economy occur in the wake of a major disaster. It has been argued, based on the aftermath of the Great Alaska Earthquake of 1964, that the rush of aid in response to a major disaster gives a community a chance to reverse a previous pattern of long-term decline.⁴³ The opportunity to rebuild on a massive scale, rationalizing the provision of public services to introduce the latest technology, could open a local economy to production possibilities that might otherwise locate elsewhere. While most case studies have shown significant long-term effects, whether positive or negative, the record also contains observations of little or no effect.⁴⁰ Overall, the case studies provide mixed evidence, at best, on local economic changes following disasters.

A major econometric study of a large national cross section of disaster events occurring during the 1960—1970 period,⁴⁴ found no long-term effects of disasters on population or housing trends. While this study has been criticized for using only population and housing units as indicators, the theoretical analysis conducted here suggests that population and housing changes could be appropriate indicators of local effects of disasters if the proper tests were performed. The same authors provide additional support for the no-effect results by conducting opinion surveys.⁴⁵ They find that natural-disaster concerns are not particularly important among public officials, many of whom might be charged with dealing with their consequences. Of course, recent occurrence of a disaster can elevate the priority of hazard/disaster concerns temporarily; however, disasters were far down the list of priorities for most officials who responded to the survey. The evidence of sensitive housing-market reaction to the announcement of earthquake risk³³ contrasts sharply with reports on the lack of long-term effects44 and the reactions of real estate market participants.³⁴

The literature on regional economic effects of natural hazards reviewed above is not based on standard economic models of an urban region. Such

models require a city to compete for capital and labor inputs against other regions.⁴⁶ It is possible to add natural-hazard considerations to the standard open-city model.⁴⁷ These models stress the effects of hazards on the supply side of a regional economy that must compete with other locations in a general equilibrium model of production.

In order to attract capital, a city must provide an expected risk-adjusted rate of return equal to that available elsewhere in the economy. Similarly, the firms in the city must be able to provide expected real wages, adjusted for differences in cost of living, which are as high as the real wages provided outside the city.

The probability of a natural-hazard event lowers expected real returns to capital and expected real wages. This inhibits regional growth. This is consistent with maximization of economic welfare in that development should not be encouraged where the probability of natural-hazard damage is high. Still, if labor and capital are extremely productive in an area, it may grow substantially even if natural-hazard events are quite likely.

Occurrence of a hazard event, such as a serious earthquake, has both immediate and long-run effects on city economic development. The immediate effect is based on destruction of capital, both industrial plant and equipment and housing. This results in lower realized rates of return on capital and lower real wages than were anticipated. Government assistance programs may raise these realized rates of return by compensating firms and households for immediate damage to capital. The relationship between payments to replace damaged capital equipment and the true measure of immediate damage (based on the difference between expected rates of return and real wages and realized rate of return and real wages) is problematical. It is possible for some individuals to be damaged substantially, through loss of expected real wages, without sustaining any capital loss that is the object of direct government compensation. Similarly, the replacement cost of business plant and equipment bears an uncertain relationship to losses in expected returns.

The long-run effects of a natural-hazard event arise because firms and workers will produce new estimates of expected returns and real wages based on experience of a disaster. Specifically, experience of a serious earthquake may cause firms to lower their expected returns to plant and equipment investment and workers to lower expected real wages, so that capital and labor are encouraged to locate outside the city.

A detailed analysis of a general-equilibrium model of a city subject to natural-hazard risks is presented in The Local Economic Effects of Natural Disasters.⁴⁸ While there are many implications of such models, among the most interesting and relevant for the discussion of serious earthquakes are the following. The one reliable indicator of overall effects on the regional economy of a natural disaster is the change in land values. Other seemingly attractive indicator variables (e.g., wages, population, and output) do not provide a reliable index of the shock to the economy generated by the event. It is possible to have an earthquake event enhance productivity by destroying outmoded capital stock and perhaps provide an opportunity to rationalize the

use of land in the rebuilding process. This result may appear shocking to many. Only a handful of economic studies have used the land-value approach. Nevertheless, this is the approach suggested by theory. A second result is that immediate effects (i.e., an inventory of capital investment destroyed and human injury) have little relationship to the overall effect on the regional economy. The long-run effects, which may be far larger than immediate effects, axe generated by changes in expectations of future productivity. Finally, a natural disaster can affect the regional economy through a number of routes, including both labor and capital markets, and a general-equilibrium economic approach is necessary to trace the outcomes.

The primary factor generating long-run effects of a serious earthquake is the change in expectations of future returns to capital investment and real wages of labor in the region. Simple announcement of a significant increase in earthquake hazards can have expectations effects as well. Thus, an understanding of economic effects of earthquakes must consider the way in which such events cause individuals to update their forecasts of future earthquake activity.

Land and housing markets in a city incorporate an adjustment to the prevailing expectation of natural-disaster frequency. The expectations hypothesis implies that, if actual disaster rates equal expectations, there should be no significant response in the city housing market, because unanticipated disasters are equal to zero. Thus, there is a need to develop an anticipated-disaster-frequency measure in order to determine if the actual disasters were more (or less) frequent than expectations, i.e., to measure unanticipated disasters. This question of expectation formation and measurement of those expectations is essential to understanding effects of earthquakes.

The expectations hypothesis regarding market responses to disasters implies that, if the frequency of earthquakes in an area during the 1980s were identical to prior expectations, then the observed disaster rate during that period would have no effect on economic activity. Unanticipated disasters are equal to zero in this case. If actual disaster experience were significantly higher (or lower) than expectations, the expectations hypothesis suggests that disaster expectations would rise, and consequent negative effects on land rents would be observed.

For example, the occurrence of three earthquakes during the 1980s in a part of California expected to have one (three) [five] floods per decade should have a negative (neutral) [positive] effect on expectations of earthquake hazard and a corresponding positive (neutral) [negative] effect on the local economy. In an area expected to have three earthquake events per decade, the danger of earthquake has already been discounted at that frequency and is reflected in both land values and levels of employment and population. As unanticipated disasters increase from-2 to 0 to +2, the local economy experiences

increasing negative effects. Note that these negative effects arise even if the community makes extraordinary efforts to mitigate disaster damage, because the expenditures for mitigation lower the expected returns to investment. Obviously, such mitigation may be a very sound economic investment if disaster expectations rise, but this merely recognizes the cost of the potential disaster event.

The expectations approach to measuring economic effects of natural disasters was tested using property value data for U.S. cities.⁴⁸ Working with flood, fire, and windstorm events, it was found that unanticipated disaster events had a negative effect on property values, while anticipated disaster events had no effect. This explains the failure of some studies to find negative effects for the total number of disaster events. The long-run effects calculated as a consequence of unanticipated disasters were far larger than estimates of immediate damage from these events.⁴⁵

The expectations approach to earthquake hazards suggests that the long-run effects of a particular serious earthquake event may be very substantial. However, these effects depend on prior expectations and the way in which expectations are updated as a result of the event. There has been little research into the way in which individuals modify their expectations of natural disasters in general and, in particular, of earthquakes. At present, there is no mechanism for assessing the nature of such expectations in different regions of the United States. Without such initial expectations information, it is difficult to anticipate the full long-run effects of a serious earthquake.

Research into the manner in which disaster expectations in general and earthquake predictions in particular are formed and the role of the government in providing an information base would be most useful. It is particularly important to determine if market responses to new information, in the form of land-value changes and mitigation efforts, are efficient. Land value studies conducted for areas experiencing disasters can allow assessment of long-run economic effects. It is not clear that disasters have permanent, long-run effects on economic growth of regions.

The following presentation is a summary of the work of Ann Butler, NeIl Doherty, Anna Kleffhe, and Howard Kunreuther—at the Wharton School. What are the appropriate policy tools to utilize as this very difficult problem

related to earthquakes is dealt with? There really is not a very good understanding of on the risks from an earthquake. There is a lot of uncertainty and ambiguity about the probabilities of an earthquake and both the primary and the secondary losses—the economic impacts on a wider level.

The role of loss-mitigation measures is one aspect of this problem. It is a tricky problem in several different dimensions. There will always be the difficulty of going from the catastrophic earthquake to the concept of expected losses and expected benefits. Following the impacts of a catastrophic earthquake, this mitigation measure may have cost only $1,000 for a private home, yet would have saved $20,000 in losses. Then one would ask the question, What is the probability of that particular quake occurring? If it is 1 in 100, you multiply 1 in 100 by a saving of $20,000 and it becomes only $200. As a result, the mitigation measure which looked very impressive after the fact may not have been so impressive if it is evaluated before the fact.

There are three areas that need to be covered: (1) the set of loss-mitigation measures for reducing quake losses; (2) preliminary results from an interactive mitigation model that the Wharton Risk and Decision Processes Center has undertaken for the Federal Emergency Management Agency (FEMA) and Dames and Moore—It utilizes benefit-cost analysis to evaluate alternative mitigation measures; and (3) recent research on the role of insurance in encouraging households to adopt the loss-reduction measures (LRMs).

A motivating question is: Should loss prevention measures be required as a condition for households purchasing insurance? The simple answer is that it depends. What assumptions are being made regarding the choice processes of individuals? How accurate are the estimates of the risk made by the homeowners who are considering the purchase coverage, as well as the insurance industry that is setting the rates?

What kinds of decision rules are utilized by individuals? We will start off with the assumption that people are rational and maximize expected utility or expected value. In reality there are many other decision rules that people use in making these decisions. That will have an enormous impact as to whether mitigation should be voluntary or a condition for insurance.

Another aspect will be, of course, the nature of the insurance program. Is it voluntary? Is it required? How are the rates going to be set? What is the level of coinsurance, deductibles, and all terms of coverage? There are a whole set of externalities associated with disasters that need to be considered in determining the role of mitigation measures. Lives lost, injuries, business interruption, irreplaceable objects, and the impact on the community of a catastrophic earthquake come into play, of course. What is the nature of disaster relief, if one is not protected if these mitigation measures are not taken? Will government come to the rescue and bail the victim out? Of course, if the victim really believe that someone is going to bail him out, then there is no reason that he should take action before the fact.

Now, a question that could then be raised is, what is the role of loss reduction measures (LRMs) in dealing with an earthquake program? There

are a number of obvious ones. It will reduce the physical damage from the earthquake, and the number of injuries and the lives lost, as well as the direct and secondary losses discussed in previous chapters. The workers' compensation, mortgage default, temporary housing, or business interruption may be spared if mitigation measures are adopted and some of the damage is prevented. It is hard to determine what mitigation will do to one's personal insurance rates, for example. Of course, it reduces the need for disaster relief.

The following is an interactive model that gives an idea as to how this problem is viewed from the point of view of expected benefits and costs. The assumptions may be tricky, but the model is very simple. Consider a particular earthquake. What is the probability of that earthquake occurring? The damage that would have occurred with and without the adoption of mitigation must be multiplied by the estimate of probability to determine the expected benefits of mitigation. If there is a very large saving from mitigation but a very low probability of a catastrophic earthquake, then the expected benefits from mitigation may be relatively small. When the benefit-cost analysis is utilized, a very small number for expected benefits may result, even though an earthquake occurs, and the reduction in damage could be enormous.

There are a number of interesting issues that affect the way one evaluates the performance of mitigation measures. One of them has to do with the discount rate. If the losses are discounted by a very large discount rate, then the expected benefits are going to be much smaller than if a smaller discount rate were utilized. Mitigation measures are paid for when they are adopted, but the benefits accrue over the length of life of the house. If the future is discounted by a very large interest rate, a relatively small benefit results from the mitigation measure if an earthquake occurs. Think, for a moment, 10 to 15 years ahead of time. Should the earthquake loss occur 50 years from now, it is going to be viewed as insignificant in present-value terms.

Another issue is the sensitivity by different stakeholders to different losses, if it is the direct losses you are talking about. What impact will the mitigation measures have on the property owner, the insurance industry, the general taxpayer, the developer, and the real estate agent? A lot of these interactions that have occurred between stakeholders is why we have a very fascinating political problem today. Certain parties are going to benefit and others are going to lose from specific programs.

As a part of the interactive model, we undertook sensitivity analyses with respect to the discount rate and the types of losses. The situation where the benefit-cost ratio will not be as attractive for any mitigation measure is when you have a very large discount rate—for example, 8 percent—and the damage is restricted to property. If the benefit-cost ratio is greater than 1, in this case, the mitigation measure is an attractive one.

On the other hand, suppose you utilize a zero discount rate over a 50-year period and you take into account secondary losses. If the benefit-cost ratio is less than 1, then this mitigation measure should not be adopted.

Now consider the case of a decision maker, like a homeowner, who has to make a decision whether or not he wants to adopt a mitigation measure.

The first assumption is that the homeowner is risk averse. The whole theory of insurance assumes that people are buying insurance because they are averse to risk. They want to avoid a catastrophic loss by paying a small premium. People are assumed to be rational, and they maximize their expected utility.

A second assumption is that the loss-mitigation measure is amortized over a period of time, 20 years. The reason for that is because there is a very large up-front cost, but there is a benefit that accrues over a long period of time. So the mitigation measure will be broken down to an annual cost, even though in reality all the money would be spent up front. The mitigation measure will impact on the magnitude of the loss and/or the probability of the loss of a given magnitude.

A mitigation measure could be thought of as reducing the damage, from $50,000 to $20,000 for a particular quake that would occur. Or it could be viewed in a different way. For example, what is the probability of a quake causing a certain amount of damage, like $50,000? If the mitigation measure, was imposed that probability might go down to zero, or it might go down to a much lower number than $50,000. A mitigation can be perceived as wither a loss-reduction measure or a probability-reduction measure or both.

It does not seem like an important assumption to make from the point of view of what will be covered on benefit-cost analysis. It makes a big difference analytically as one begins to look at the model. Benefit-cost analysis will be discussed later in this presentation. It is something that has come out to be a rather important element in terms of judging what is good and what is not so good from the point of view of benefits.

The following four considerations will guide the analysis:

1. Myopic behavior. It is recognized with mitigation measures that cost is borne once and benefits accrue over time. Immediate costs can be focused on, so if asked whether one wants to adopt a mitigation measure or not, one would say this measure will cost me $500. Someone says, you are going to live in that house for 20 years. Why not discount it? why not look at all the benefits over this period of time? The response would be the same, the cost is $500. That is what must be paid. There is no interest in what is going to happen 10 years from now. If people behave this way, then a lot of the assumptions that have been made on the expected benefits may have to be modified.

2. Misperception of the probability. This earthquake is not going to happen to me. Why should it happen to me? I am living in an area, and we had our quake a year ago in San Francisco. I am certainly not going to think about an earthquake happening in the next 10 years. If it is thought that the probability is extraordinarily low, mitigation is not going to be an appealing tool. Why

worry about making an improvement on my house when the earthquake is never going to happen?

3. Availability of disaster assistance. If disaster relief funds are available from the federal government, why improve the house? In the past, the federal government has come to the rescue, and people may expect them to do so again if there is some kind of catastrophic event.

4. Impact of insurance coverage on adoption of mitigation measures. The first factor is whether or not one has risk-based rates or rates that do not change as a function of mitigation. The second factor has to do with how insurance impacts on risk aversion. If insurance is purchased, interestingly enough, there is less incentive to mitigate, everything else being equal. If insurance is purchased, the purchaser is protected with a small premium against a very large loss. What is being done through mitigation? The purchaser is protected with a small payment against a larger loss. By taking out insurance, the aversion to risk has already been reduced. Therefore, mitigation in and of itself is a less attractive option than if the insurance had not been taken out.

For example, there is an earthquake of intensity 8.0, with the probability of .0181. The home value is $100,000, and the damage without mitigation is $11,000. The damage with mitigation is $4,700. The cost of mitigation is amortized on an annual basis to obtain $111 a year, and the benefit of mitigation is $114 per year amortized, so the benefit-cost ratio is 1.03 on these figures. The premium on an insurance policy will be determined by the probability of the earthquake. Notice that the focus here is on just one catastrophic earthquake.

Two different cases present some qualitative results:

• In the first case everything is accurately estimated. The probability is accurately estimated by the individual, and the question is, what is the benefit of taking out mitigation measures or not taking out mitigation measures? One benefit is no insurance, the other is full insurance. Mitigation comes out best whether one has no insurance or full insurance as in this case. The time that mitigation is most likely taken is when there is no insurance.

• The second case is one where there is misperception of probability, and it is still thought that there is a chance of an earthquake, with the higher probability of .06. In this particular case, when there is no insurance, no mitigation is better than mitigation. One would say, why mitigate the problem when its probability is so small? I would rather take the loss if I am going to measure it that way. If you are forced to buy insurance, then it turns out that mitigation is better than no mitigation if you have risk-based rates. The risk-based rates are based on the true probability rather than the perceived probability, and the benefits received from mitigating are going to be greater than the cost.

Several policy questions can be explored. What factors lead individuals to voluntarily adopt the LRMs? That is an interesting behavioral question

where there is some insight in terms of what people are likely to do. Some of this analysis may shed some light on what people could be encouraged to do, but a lot of this has got to do with how people perceive the probabilities, how they perceive the losses, what incentives can encourage people to take action, and when will regulation be needed.

There are some interesting questions in terms of differences between floods and earthquakes. Currently in Congress there are some questions as to whether earthquakes should be treated in the same way as floods have been. There may be some similarities, but there are also large differences between floods and earthquakes. One has to figure out how that is going to impact on the proposed program. The carrot-and-stick approach is appealing, but the question is, what kind of a carrot and what kind of a stick should be used in this situation?

QUESTION: Dr. Kunreuther, how does the individual's life expectancy comes into the picture? Why worry about what happens 100 years from now

DR. KUNREUTHER: That is an interesting question, but is it a probability of .01 or is it that the earthquake is 100 years from now? As we know, it makes a real difference in terms of how one interprets it; and, certainly, if one thinks that it is 100 years from now, then you are right. At that point the probability is zero for the next 99 years and 1 for the hundredth year, if you really interpret ''100 years from now'' in that kind of a context. Of course, nothing matters then, just live happily ever after.

QUESTION: Your least favorite point is that individuals behave rationally. Do you find they behave more rationally when it comes to material things—such as a house, a car, physical damage—and less rationally when it becomes their perception of personal risk? I think of this in relation to the nuclear power debate, where the material risks are relatively negligible but the personal, perceived risks are enormous.

DR. KUNREUTHER: Right. Let me give you an overly simplistic answer with lots of grayness. There seem to be two kinds of events that individuals deal with, and they are dealt with differently. They are the events where we say, "It cannot happen to me." This occurs partly because they are voluntary; it is our decision. These are the kinds of events that we really prefer not to worry about because if this Pandora's box of horrors is opened, we start worrying about more and more things. With respect to natural disasters in general, no one is to blame. There is a tendency to think about these events by saying that the probability is very low or I am not going to worry about the consequences.

The contrast with the nuclear power, hazardous waste, and NIMBY problems is extremely important, because here we almost take the opposite tack. We say, "It can happen to me;" and, even if the experts are going to tell us that the probabilities of these events are extraordinarily low, we do not

necessarily trust the experts. There is a lot of ambiguity associated with these issues and, most important, we can blame someone else for the consequences. Therefore, we do not want it in our backyard, even if people are going to say that there are real benefits.

I think one of the challenges in this whole area is to start categorizing human behavior as we look across these events. One particularly intriguing event that has not been fully studied is radon. Radon is a case where you cannot blame anyone for the damage, and yet it may be in your own house. It is not at all clear what kind of decisions people will make in terms of mitigation, for example, because they should be willing to take some action to get rid of the radon; yet everyone is saying, "I do not want to know whether my house has radon. I do not want to have a test. I do not want to have anything to do with it simply because if I know, I have to do something, and I am not sure that I want to know."

QUESTION: When considering the process of the lack of attention to mitigation in public buildings, shouldn't the function of the building be taken into account? For example, a warehouse and a school have very different functions.

DR. KUNREUTHER: That was one particular example and a particular kind of mitigation measure. We should not in any way generalize from this one case of one public building, but the point you are making is very important. It was just an illustration that there could be some mitigation measures in some situations which you might not want to undertake. What should come out in any analysis is, that a public building has lots of secondary impacts, and there are lots of benefits when all of those things are examined. In most cases, mitigation measures probably will be very beneficial.

QUESTION: Dr. Yezer, how does this relate to your concept of a public building and its role in secondary impacts?

DR. YEZER: Ultimately, we pick up the public buildings and the land value. Presumably if a park is destroyed, the land value around the park obviously is reduced. That is how it is incorporated in the model which I presented (Chapter 4).

In addition, the issue of how people react to probabilities of disasters is interesting. People assume that individuals do not react to probabilities well. If that is true, they should overestimate probabilities of future disasters right after they have just had one and underestimate them when they have not had one for a period of time. That would suggest that there is gain to be made by buying and selling houses, that firms should be ready to rush into areas that have had disasters and buy up all the houses at very low prices. I have heard of some of that behavior; but while people may say all sorts of irrational things about their expectations, they refuse to conduct a fire sale of their home. At least the market has not found them willing to do so, therefore you have to watch out for the difference between what people say and the actual behavior of the individuals in moving or in selling an asset.

QUESTION: Dr. Yezer, my question relates to economics. Even when the probability of a disaster increases, there is no apparent decrease in property values.

DR. YEZER: I was trying to illustrate in that particular case, you can have the same effect. You have the same exact physical event. In one case, you observe a lot of buildings falling down, and in another case you do not observe any falling down, because you spent all the money on mitigation. Of course, the key is that the expensive mitigation has been capitalized in the land value. In fact, you would have an unambiguous indicator of the effects of earthquakes if you focused on the right variable—land values.

It is interesting that everybody wants people to mitigate, and they are focusing on all of the observed dollar flows that they can find if there is a disaster and there is no mitigation. That is a perfectly sensible thing to do, but one must realize—just as in the case of highway accidents—that we do not want to reduce fatalities to zero, that we really do not want to have the country spend so much of its GNP on mitigation that we reduce hazard damages to zero.

QUESTION: My question is related more to perception of risk and how large a factor disaster risk is in market forces or real estate values.

DR. YEZER: Well, I would bet that property values do not fall by more than 1 or 2 percent as a result of a Loma Prieta-size earthquake. But even 1 or 2 percent of $50 billion in land values in the Bay Area is a nice piece of change. As a matter of fact, using the sort of approach I have advocated, you get much bigger effects of unanticipated disasters than you would ever get by counting the buildings that actually fall down.

QUESTION: What is the role of insurance in mitigating secondary impacts. Specifically, what were the economic effects of Loma Prieta and what role did insurance play versus the economic impacts of, say, Hurricane Hugo and the role of insurance in that economy.

DR. KUNREUTHER: That is a very interesting question because the question is really, what kind of a policy is one talking about from an insurance point of view? If you are referring to an earthquake policy that is related to property damage, that is one kind of an insurance policy. If you are referring to a business-interruption-risk policy, that is another kind of a policy. Certainly we could consult with people who are much more knowledgeable than I am about some of the details with respect to various insurance. The question that we would raise would be, are you attributing essentially the losses that come out of the quake to the right kind of policy or are you subsidizing across policies? For example, if it is life insurance, a quake policy is not really protecting yourself. By mitigating, for example, you may be able to save lives. So in some sense, through mitigation, you are cross-subsidizing life insurance and likewise with business-interruption risk, if that happens to be a separate policy.

There are two issues. The first issue is: What is the function of insurance? This is a question of whether something is insurable. Personally, I would want to turn to the people who really are marketing these policies to

make some comments on that issue. The second issue is: Are the rates really reflecting the risks or is there some kind of cross-subsidization across risks?

QUESTION: This was a question for Dr. Yezer. As I understand it, the indicator in your model is changes in the values of land. If that is so, then how would you explain that around 25 years ago, when identification of the earthquake threat significantly increased, there was a trend of increasing land values in Los Angeles, San Francisco, and Tokyo, which are the highest land values in the world?

DR. YEZER: Well, other things have been happening in those areas, and you have to try statistically to hold all the rest of that constant. I can get a measure of the rate of appreciation in values of houses that was anticipated before a hazard effect, then I observe what happened afterwards and try and explain the difference. Expected appreciation rates before any hazard events in California, they are very high. High rates of appreciation occur in high-risk areas, but the only question is the partial impact of unanticipated-hazard events on these property values.

There are a lot of anticipated-hazard events in California. For example, everybody knows that if you have a heavy rain, a lot of houses are going to slide. And, when you have a heavy rain periodically and those houses slide, that does not really have any effect, because that is an anticipated disaster.

QUESTION: I am just wondering about this issue of trying to assess losses and damages. The smaller the region defined, the more the impact. For example, in the Marina District there was an incredible impact. In San Francisco, houses which are on bedrock are appreciating, and the value of land and property which are not on bedrock has gone down. There was a very significant microdifference in land value, in land prices. That is a very important issue.

DR. YEZER: You have to take this across a labor market area. Charles Scawthorn wrote a paper on what is called the "tilt effect" in the Journal of Urban Economics. If one area of Tokyo has an earthquake, then overall land values will go down; but in some parts of the city that did not have damage, land values will go up. He has modeled this effect by taking a full labormarket area in order to account for this tilt. If a flood occurs, then the property values in low-lying areas will tend to go down, and the high-lying areas will tend to go up. But overall, land prices will go down if there is an unanticipated disaster.

QUESTION: However, in your model, it seems you are assuming that every structure is going to have a loss at some point. I think that over time-say a 100-year period—there will still be houses or buildings. Was that taken into account in your process?

DR. KUNREUTHER: I presented a representative structure. As you begin to start doing this on a more systematic basis by going across structures, you would obviously have to tailor the analysis to the individual buildings. The idea basically was that there was a probability of a particular quake occurring, and then one would ask: If that quake occurred, what is the likely damage to a particular structure? As you move that structure away from the fault line,

if it were a wood-frame house rather than a concrete block, you would have different kinds of damages that would take place for any kind of an earthquake. It is really a generic treatment, much in the spirit of probabilities toward losses. To the extent that you were able to look at the characteristics of that structure—and, of course, that is the reason why loss-mitigation measures get considered in the first place—you might be able to say that the damage that would be sustained by that structure would be very, very small, even for a very large earthquake.

QUESTION: If you assume a value of $100,000 for a house, the societal cost would require that we account for the whole community, and not just for individual people.

DR. KUNREUTHER: I think that is one of the interesting questions. When thinking about mitigation or kinds of policy tools, we must consider our society in various situations. We have a very hard time trying to measure the specific characteristics of these structures. This is similar to the problem we face in the insurance industry in general. Since you cannot get information on every little characteristic of each person, rates are set for a particular class. The individual may feel discriminated against because they say, "I know I am better than the average driver." Yet the principle exists that you have to establish a policy upon which a rate is based.

With respect to earthquake insurance, it would be preferable to tailor rates on the basis of knowing about each specific structure. But if it is going to cost too much to do it, we may not be able to base rates for every individual structure.

DR. YEZER: There really is an interaction between what Dr. Kunreuther is doing and what I am doing. If you get risk-based insurance premiums, then people have a basis for judgment about allocating their resources, both capital and labor. You can provide a tremendous amount of information about risk from the insurance premium. Providing information to people allows them to respond efficiently. Risk-based insurance premiums could give you a lot of information without having to go out and read books about geology and earthquake probabilities.

DR. KUNREUTHER: We also have in common a Chicago model as a part of our orientations, recognizing that it is limited in terms of what it can do. Also recognizing that things like incentive systems are voluntary and more desirable than requirements. But there may be situations where you have to have requirements. Risk-based premiums would be an example of a real incentive to give people information, an instance where you do not have to make the requirement. Will they behave in a way that takes into account that information? That may be another question.

QUESTION: There is a cost of risk-hazard mitigation. Who has to pay that cost?

DR. YEZER: Those who benefit from mitigation should pay. Transaction costs of enforcing mitigation measures is an important issue. Who pays for this? Is there a bill of sale or some kind of a contractor's bill that can be taken at face value, to prove that someone has done something to make their

building safer? How can the effectiveness of mitigation expenditures be assessed? What are the mechanisms that will be used, and how will they be implemented?

QUESTION: Is your selection of 20 years for this kind of mitigation completely arbitrary? Why would it not be the life of the house?

DR. KUNREUTHER: Purely arbitrary, because we like the number 20. I think it is important that you bring it up, because we had to pick a figure, so we picked 20. We could have picked 50, we could have picked 80.