Overview of Economic Research on Earthquake Consequences
This session is devoted to a state-of-the-art overview of economic research efforts that have focused on the consequences of major earthquakes. The purpose is to identify what is known about these consequences, both theoretically and empirically. What are the strengths and weaknesses of these efforts to date, and what gaps remain to be filled?
Professor Harold Cochrane, a professor in the Department of Economics at Colorado State University, will present this overview. Dr. Cochrane has extensive research and advisory experience in environmental issues and natural disasters, especially with respect to the distributive effects and economic consequences of such events. He has been involved in three recent NSF-sponsored projects investigating a variety of economic issues on natural-disaster losses.
PRESENTATION OF HAROLD COCHRANE
The state of the art of economic research on the consequences of a catastrophic earthquake is both abysmal and elegant, the former because data are seldom collected in a form consistent with accepted loss-accounting principles, and the latter for the level of mathematical sophistication attempted. This overview will be devoted to the often forgotten but essential principles for assessing impacts and will avoid technicalities, serving instead to highlight a broad range of issues which go well beyond those included in conventional damage assessments.
The issues presented are the product of several NSF-sponsored projects to develop a guidebook for practitioners conducting damage assessments. The book is entitled, Damage Handbook: A Uniform Framework and Measurement Guidelines for Damages from Natural and Related Man-made Hazards. The need for such a handbook was identified by Eleonora Sabadell. Its application to the Loma Prieta earthquake was made possible by William Anderson at the Foundation.
Earthquakes produce physical consequences. Tens of thousands of people may be affected, either directly or indirectly. Secondary events such as fire, landslides, and dam failures may widen the scope of damage, possibly including injury to fragile ecosystems. However, from the economist's viewpoint, the mere description of physical consequences is just the beginning of a damage assessment. The most challenging task, as Professor Kling so ably presented in Chapter 2, is that of valuation. This overview will not tackle the valuation problem in any depth, since it would only repeat many of the points raised in Professor Kling's presentation. It will focus, instead, on a set of
principles, distilled from an integrated accounting framework. Each will be discussed in turn and will be applied to the Loma Prieta earthquake. Finally, the impact of the Great Tokyo Earthquake of 1923 on the U.S. economy will be speculated upon.
Useful Concepts for Identifying and Measuring Economic Damages
Catastrophic earthquakes typically cultivate images of death and devastation, which might be extrapolated to include a litany of economic impacts. Although each clearly reflects a different aspect of the event, they are not additive. Double-counting and miscounting losses is a problem so severe that little credence can be placed on current damage estimates.
The simple relationships of accounting identities and budget constraints, which govern a regional economy, are the fundamental building blocks from which economic damages are isolated and measured.
Income Accounting and Damage Assessment
In its simplest form, a regional economy consists of a producing sector, consumers, and government. The producers employ members of the household to work with the existing plant and equipment. Incomes are paid to the laborers and those who have capitalized the firm (i.e., the firm's financiers) are entitled to wages, interest, and dividends. The demand for products provides the firm's owners with the incentive to continue producing, which also means that the workforce, plant, and equipment will continue to be utilized.
Not surprisingly, the demand for products stems from the incomes earned. The presence of government causes some spendable income to be diverted from households, but as Figure 3-1 shows, government purchases and payments produce a new set of demands, which may or may not be equivalent to what households would have done in the absence of government.
Without doubt, this is the simplest characterization of economic activity. However, even at this abstract level, several important principles are worth noting. First, the level of economic activity can be measured by counting expenditures, or incomes, but not both. Income to the firm's owners, its workforce, and the firm's financiers must be equivalent to value of the products produced. This is because the price of a product reflects all the costs incurred in its creation, which in this case is the sum of wages, interest, and profits. This simple result provides an important loss-accounting guide: damage assessments could focus on incomes lost or spending lost, but not both. Either should yield the same result.
This schematic of the economy also explains why lost sales taxes and the local government services such taxes support should not be added to lost income. Double-counting is involved here as well, albeit in a more subtle form.
Since the primary source of government revenue is traceable to income (sales and income taxes) and property value (property taxes), a disaster-induced reduction in these values implies a proportionate reduction in government revenue, with one minor exception, the latter is derived from the former. It is therefore incorrect to add loss of government revenue (and the associated services) to property and income losses. To do so would count losses twice.
This description is, of course, oversimplified, since regional economies axe embedded in a much broader and more complicated matrix of economic forces. For example, households may attempt to spread their losses over time by liquidating real or financial assets or going deeply into debt. Insurance payments, philanthropy, and federal aid will tend to soften regional losses, thereby masking the real effects of a disaster. These possibilities illustrate the pitfalls inherent in loss measurement and further underscore the need for a carefully developed framework.
Direct Versus Indirect Economic Impacts
The so-called indirect effects of disaster are often confused with the direct impacts. Direct impact refers to the loss of plant and equipment which stems directly from damages sustained in the event plus any associated loss of employment. These losses may produce supply bottlenecks, which may produce a ripple effect, inducing layoffs in related but undamaged industries. Such effects have not been widely observed, although some have speculated that the loss of semiconductor production in Santa Clara would produce such effects. Even this scenario seems farfetched today, given the recent emergence of Japan as a major world supplier. It may be that the Great Santa Clara Earthquake simply eliminates the United States as a chief force in the computer chip industry, a trend which has been clearly evident over the past decade.
Reductions in household incomes resulting from layoffs, bankruptcies, and bad loans would produce a separate set of economic effects, referred to as induced or multiplier effects. These too axe lumped under the category indirect effects. And, these too have been difficult to detect in Loma Prieta and Hurricane Hugo. The decline in tourism suffered by the barrier-island resorts of Isle of Palms and Sullivans and Folly islands was more than compensated by the huge injection of insurance claim money into the state. By all accounts, South Carolina is experiencing an economic boom of sizable proportions. The same cannot be said for the Bay Area region, since the amount of outside resources, insurance, or federal aid funnelled to the region has been comparatively meager. Even so, the secondary economic effects of the earthquake axe nearly impossible to detect.
There are two reasons why secondary losses are so difficult to detect: (1) the economy is more resilient than most economists like to believe, and (2) the effects of disaster axe shifted to other regions or to another time period (possibly to other generations). In a world of federal budget limitations,
increased aid to help earthquake victims means either tax increases for the general population or a reduction in spending somewhere else in the nation. Even if lawmakers can successfully negotiate an increase in the budget deficit by keeping the aid off the budget, as they did with the savings and loan bailout, the nation is still left with the tax liability which the debt represents. In this instance, the costs of reconstructing an earthquake-torn city are left to our children. Multiplier effects and related secondary losses are, therefore, either pushed onto other regions or shifted in time. They are not eliminated.
Real Versus Financial Effects
Many have speculated about the real effects of disaster, but these effects have been seldom observed. There are several instances where a natural disaster has triggered a depression. Two of these instances are: (1) the Dust Bowl period of the 1930s in the United States, and (2) the Managua earthquake. But even in these instances, the natural event was accompanied by man-made events, the stock market crash and a political revolution, respectively. Counter examples are clearly more plentiful.
So, why do some disasters trigger or at least accompany severe economic contractions and others do not? The answer lies in the underlying strength of the economies affected. The United States has learned from Managua, Nicaragua, and Johnstown, Pennsylvania, where it has been observed that disasters tend to accelerate ongoing economic and social processes that were working prior to the event. Failing economies experience a sudden collapse, whereas robust economies experience a boom. There are sound economic reasons for these observations which will be addressed later in this presentation.
The financial effects of disaster are simply repercussions of the disaster on the stock and bond markets. There are many reasons why stock and bond prices move, but clearly psychology or the state of expectations plays an important role. Ignoring this important issue temporarily, the occurrence of a large disaster may or may not impact these markets. Clearly, the value of corporations directly impacted by the disaster will suffer, for they have lost productive capital. And, to the extent that the market value of the corporations affected reflects the valuation of income streams these companies are capable of generating, their stock prices should decline. In a rational world, these markets would decline by the value of the capital lost.
The total loss from Hurricane Hugo and the Loma Prieta earthquake combined amounted to under $10 billion, nine-tenths of which was sustained by homeowners. The value of the equity market in this country is over $2 trillion. This means that these disasters may have produced a .005 percent change in the capitalized value of corporate America, an insignificant amount to register on the stock exchange. This is not to say that some corporations, such as insurance carriers and resort companies in South Carolina, were not impacted. But overall, the effects were negligible, particularly in contrast to
the wild swings in the New York and Tokyo stock exchanges, which caused nearly $2 trillion in paper losses combined.
The following seven loss-measurement principles reflect the conceptual orientation just presented.
Measure Loss With and Without the Earthquake, not Before and After. This is one of the clearest and simplest economic principles guiding impact assessments. It is also one which is commonly violated, however. Loss studies are, of necessity, conducted over long time intervals, during which economic pressures unrelated to the disaster can mount. Because of this, it is possible to conclude that a disaster produced an economic change, which more rightly should be attributed to unrelated but correlated factors. U.S. Steel's decision to shut down the Johnstown mill after a devastating flood in 1965 was linked to a chronic decline in the plant's profitability, rather than the sudden onset of damage. The event simply provided management the excuse to terminate operations, a decision which would have eventually been made with or without the disaster. Under such circumstances, it would be incorrect to attribute loss of jobs and the accompanying economic downturn to the flood.
Do Not Double-Count Impacts. The accounting framework discussed earlier pointed out several ways in which losses could be double-counted:
damage to assets versus flows—The destruction of productive capital reduces the region's flow of income. Both cannot be counted.
expenditures versus incomes—They are related. This applies to households, businesses, and governments.
financial versus real—Financial effects mirror the real.
property values and damage—Risk of earthquake damages should cause real estate values to decline. Both cannot be counted as a loss.
Damage Assessments Should Not Include Land and Depreciation. It is important to distinguish between damage to structures and a possible reduction in the value of building sites. This is dearly a nontrivial point. In parts of California such as San Jose, the value of real estate is driven mostly by location, with building improvements contributing only a small percentage of the property's overall value. Occasionally, damage assessments incorrectly utilize the total market value of a house in deriving estimates of earthquake losses. This of course is inaccurate, since the site still has value. In fact, some unreinforced masonry buildings may have negative value relative to that which an alternative structure or land use might bring. This is particularly true when rent controls and tenant laws serve to prolong the use of unsafe buildings. Failure to separate equity in the structure versus equity in the platted site may
explain why homeowner equity is an insignificant variable in the decision to purchase earthquake insurance.
Expected Losses Should be Derived from a Dynamic Building Stock. The inventory of structurally unsafe buildings will change with the occurrence of events. A moderate earthquake in the range of Richter 6 to 7 would render many unreinforced buildings uninhabitable. Condemnation (red-tagging) of such buildings and their subsequent demolition would reduce the stock of hazardous buildings subject to failure in the event of a catastrophic shock. By assuming that the number of hazardous buildings remains constant over time, as is typically the case, the benefit of seismic improvements may be overstated.
Mitigation Policy Should be Based on Avoidable, Not Total, Loss. Loss management should emphasize the trade-offs involved in purchasing additional seismic safety. All too often, policy-relevant losses and total losses are confused. This tends to skew public priorities irrationally. Events which evoke images of catastrophic damage often serve to fuel political rhetoric, but seldom illuminate those strategies which are economically efficient. For example, the widespread destruction Hurricane Hugo's winds wrought on the Francis Marion National Forest resulted in untold recreational losses, in addition to lost timber. However, from a mitigation standpoint, the value of destroyed timber and recreation days lost is of little relevance. One must ask whether the event could have been avoided at reasonable expense. The answer is no!
Federal Priorities Should be Based on National, Not Regional, Impacts. Loss studies are conducted for a variety of reasons, some of which are self serving and some not. Clearly, regions impacted by disaster will find it in their best interest to shade facts in the hope of triggering a disaster declaration. Although quite understandable, Dr. Cochrane is not interested in such practices. This paper focuses instead on losses viewed from a national perspective, and emphasizes in particular the use of loss data in the development of economically efficient strategies for coping with earthquake consequences.
Count Secondary Losses. With this in mind, one must exercise great care in the interpretation of loss data. The provision of federal aid, the receipt of insurance monies, or even bearing the loss by paying for repairs out of pocket will temporarily stimulate the affected economy, thereby masking secondary impacts. Transfers serve to simply shift employment changes to another point in time or to other regions. Therefore, although secondary employment effects have not been observed, they should be estimated and counted.
Observations about Loma Prieta
This overview reflects a set of principles which are commonly accepted, at least by academic economists. However, as in the case of Loma Prieta, they are often ignored in practice.
Damage estimates persist at $5 billion to $10 billion. This represents less than I percent of the region's capital stock. To put this into perspective, the loss is less than the annual rate of depreciation, or the capital loss resulting from a 0.1 percentage point change in the long-run rate of interest. It appears that secondary impacts may amount to no more than $2 billion, and may more likely prove to be half that figure. This too is a relatively small amount, less than i percent of the region's annual income.
In some cases, these estimates axe made within 48 hours of the disaster and not subsequently revised. Except for calls from economists trying to recalibrate national economic models to accommodate the effects of the earthquake, news reporters were the most frequent inquirers about losses.
The following observations about earthquake losses were gleaned from Bay Area press clippings, discussion papers, and interviews with decision makers. They are labeled as misconceptions because they violate one or more principles of loss measurement as put forward in this paper.
Lost economic growth should not be considered a cost of the earthquake . The Federal Reserve Bank in San Francisco indicated that there might be long-term-losses if prospective businesses and residents decide not to come to the Bay Area. This is not necessarily a cost. It might reflect new information about the true costs of locating in the San Francisco area.
The negative effects on employment should be tracked separately from the stimulative effects of postdisaster reconstruction. Economic disruption stemming from the earthquake was masked by outside aid (e.g., disaster unemployment assistance, Small Business Administration loans, and state and federal aid to governments). Reconstruction financed by borrowing and insurance settlements also tended to dampen secondary impacts.
Damages from the earthquake are not equivalent to the market value of the property. Even if the entire structure is destroyed, the site still has value. Be sure to account for depreciation. A new structure has a longer useful life and is safer.
Accounting stance is important. Economic impacts were uneven. Bookings at San Francisco hotels declined, while business in San Jose was unexpectedly strong.
Lost sales is a poor indicator of loss. Shopkeepers tended to report lost sales as the cost of disaster, rather than lost value added. Sales include the value of goods imported into the region, and which could be sold elsewhere.
The costs of unemployment should not be added to lost revenues. Wages are reflected in sales revenue. If it is claimed that hotel occupancy rates dropped by 10 percent, then hotel revenue is reduced by 10 percent (assuming that the rate per room remains unchanged). To then add the cost of unemployed hotel workers to the lost revenue involves double-counting; room rates include all labor costs.
Lost leisure should be included as a cost of disaster. Some analysts speculate that productivity losses were held to a minimum by a labor force
willing to endure lengthy commutes to circumvent damaged roadways and bridges. However, longer commutes imply lost leisure, which has its own costs.
Retrofitting to make freeways and structures safer is not a cost of the earthquake. The temporary repairs of $60 million to the Embarcadero, I-280, and U.S. 101 are attributable to the earthquake. (The $1.7 billion worth of seismic design changes is not.) Since the design changes reflected in higher construction costs are not a product of the earthquake, they should not be counted as a cost of the earthquake. Technically, it is appropriate to count only the cost of restoring the freeways to their prevent condition.
A decline in sales and income tax revenue is not necessarily a disaster-related loss. If damages include lost income, they also implicitly include lost tax revenue. Taxes are a function of income and spending. To include both involves double-counting.
History may be a Poor Guide—the Great Tokyo Earthquake, 1923
Much of what has been covered reflects accepted microeconomic principles. At the macroeconomic level, however, there is much less consensus about how an economy is likely to respond in the event of a large earthquake. As a result, forecasting losses at the national or international level is still a highly subjective undertaking. There are several reasons for this. Most important, disaster scenarios typically tend to reflect the economic contraction observed in 1929. Whether such a sequence of events could be repeated in the 1990s is for some reason never addressed. Second, economists tend to focus on U.S. disasters. By so doing, they are failing to recognize that the U.S. economy is now embedded in a much larger financial system, which could be destabilized by earthquakes either here or abroad. (As will become evident in the remarks which follow, I am extremely critical of the state of the art when it comes to macroeconomic loss assessments.)
The organizers of the conference wanted the presenters to restrict their presentations to domestic events. However, the Coming Great Tokyo Earthquake provided a number of interesting possibilities for discussing the previously mentioned shortcomings. The following scenario will stimulate discussion as to whether the Richter 8.2 San Andreas earthquake is still the truly catastrophic event for which plans must be made.
A recurrence of a 1906-like event in the Bay Area might cost $50 to $60 billion. Would this be large enough to create serious macroeconomic effects? Maybe not. What about a Richter 8.2 in downtown Tokyo? The cost of such an event has been estimated by Japanese economists to be $600 billion. What would be the implications for the Japanese and the U.S. economies? There can conceivably be two scenarios, one fairly gloomy and one which produces a brighter conclusion. It is not known which is more likely to occur, but the more positive scenario is intriguing.
At the time of the event, the Japanese economy is likely to be the size of the U.S. economy, approximately $4,000 billion GNP. Currently Japan's trade surplus with the United States is approximately $50 billion annually, mostly in real estate, T-bills, and select small companies. Japanese households could alone provide enough savings at the time of the event to pay for the cost of rebuilding Tokyo. No U.S. assets would need to be liquidated.
Possible Financial Repercussions on the U.S. Economy
This conventional scenario follows a pattern similar to that observed after the 1929 stock market crash. The triggering mechanism is different, but the results are identical. According to Japanese economists, reconstruction costs will be financed by a massive sell-off of U.S. Treasury bills. U.S. bond prices plummet, driving interest rates up 5 percent. The sudden rise in U.S. interest rates depresses home buying and capital investments, which produces a deep recession, paralleling the recession experienced between 1929 and 1936. Japan, on the other hand, experiences rapid growth, fueled by internal, not external, demand. Japan emerges as a super economic power, while the U.S. economy languishes.
The Great Tokyo Earthquake will strengthen both countries. First, the Japanese economists have overstated the magnitude of the event. They have probably included land values in their estimates and other impediments which have driven up the cost of Tokyo real estate. Land will not be destroyed, only structures and infrastructure. The event will be costly, but as pointed out above, the Japanese can muster enough savings to rebuild Tokyo without selling Treasury bills. The interest-rate differential will be attractive enough to induce them to hold onto these assets. Even if they are sold, the interestrate effects are likely to be smaller than the 5 percent increase some have forecast.
The event will provide a reason to implement the recently signed trade agreement, which emphasizes the opening up of construction markets to U.S. firms. It is unlikely that under normal circumstances this would occur voluntarily. Interest rates will likely rise in the United States, but that is a normal economic response. We might view Japanese investments in our T-bills as their form of disaster insurance. The disaster will stimulate demand for U.S. building materials, resources, and services. Building in the United States will be temporarily curtailed, but this should not create significant problems, since many regions of the United States are experiencing an oversupply of office space (e.g., Houston and Denver).
In short, the disaster might serve to rectify the trade imbalances which have dominated public concern for the past decade.
Some may question whether this latter scenario is overly optimistic. In considering recent events such as the unification of Germany, the New York and Tokyo stock market crashes, and the savings and loan crisis, it can be concluded that the U.S. economy is more resilient than many observers would
have us believe. For this reason, it is probable that some postdisaster projections about the performance of the economy have been overly pessimistic.
Useful Results can be Obtained from Simple Approaches
Loss estimation is imprecise, based on an incomplete and erroneous conceptual foundation, and continues to rely on hastily collected and inaccurate data. Data for many loss categories presented in this overview are simply unavailable. The purpose of loss studies is all too often politically motivated. There are exceptions, such as FEMA's and San Francisco's well-planned attempt to deal with its stock of unreinforced masonry buildings. Methods for valuing impacts must be simple. Fewer than 10 percent of those who profess to use loss data are trained in economics, and even fewer are familiar with the principles of loss measurement as promoted in the Water Resources Council's Principles and Guidelines. Despite these impediments, a simple and internally consistent set of principles can most likely be set down to assist in the presentation of earthquake losses.
It is suspected that a correct accounting of losses would show a level dramatically different than that which has been gleaned from recent events in San Francisco and South Carolina. There are several obstacles which inhibit the profession from making inroads, the most intractable of which is its stubborn attachment to outmoded views of financial and economic systems. It almost appears as if the stock market crash of 1929, and the ensuing deep depression, has left an indelible mark on the discipline's orientation. All potentially destabilizing events tend to be cast in the 1929 mold. ''The Great Tokyo Earthquake will produce panic selling on Wall Street.'' The rest of the scenario is identical to that of 1929 and 1936. "The Great Santa Clara Earthquake will produce a shortage of microprocessors, which will bring the nation's capacity to process and disseminate information to a halt." "The economic consequences will be as catastrophic as, if not more catastrophic than, that produced by the oil embargo of 1974."
Clearly, institutions and the economic realities of the 1990s have changed such that the mechanisms for transmitting economic shocks have been altered in ways we can only imagine. The U.S. economy is firmly embedded in the world economy; this more than anything else has changed the very nature of the catastrophic event.
GENERAL DISCUSSION OF CHAPTER 3
QUESTION: The Japanese studies generate a much larger loss possibility than Europe. They are pretty low actually. They were way up in the billions.
DR. COCHRANE: More than $600 billion?
QUESTION: It was $655 billion.
DR. COCHRANE: We are close enough, but it could be argued that these estimates might include land value. I must admit that I am unsure how the Japanese performed their analysis. However, I tend to be skeptical. My guess is that reconstruction activity would produce some positive spin-offs. Tokyo has grown in ways that have overtaxed the city's infrastructure. As far as the $600 billion estimate is concerned, it is startlingly large.
This type of event conjures up images, albeit somewhat dated, of economic and financial collapse. Those that forecast such dire consequences anticipate an event that triggers a massive sell-off on stocks and bonds, and that produces a rise in interest rates, declining investment, and lowered aggregate demand, all of which serve to induce a rise in the U.S. unemployment rate. Such projections might conclude with U.S. losses amounting to $1 trillion. This type of thinking, in my view, is unimaginative. I agree that this sequence of events did emerge after the stock market crash of 1929. But, it did not in October 1987, when the collapse produced approximately $500 billion to $800 billion in paper losses. Clearly, something had changed between 1929 and 1987. The economy is no longer responding as it once did. It is this change in response that should attract our attention, not a replay of the 1929 scenario. At this stage, we probably do not have a good understanding of the loss-transmission mechanism. But, in my opinion, the simplistic projections I have seen appear to be overly pessimistic. They overstate the economic impacts.
QUESTION: Double-counting is a real problem, but there are regional issues that must be considered. In order to understand and do something about some of these problems, it is important to measure things which—from an overall point of view—could be involved in double-counting. There is a real need to study that and also to study distribution practices. When we are doing an overall assessment of the disaster- or hazard-mitigation policies, we do one kind of analysis. But for regional and local policies or state-level policies, we have to do other kinds.
DR. COCHRANE: If you are a regional economist, you had better use an accounting stance that reflects your client's interests. On the other hand, since these presentations on national priorities, a national perspective must be adopted. Some of the issues talked about earlier illustrate this difference in perspectives. For example, negative employment effects felt in the disasterstricken region may be offset by positive effects elsewhere. In my view, a national—as opposed to a regional—perspective is appropriate.