Economic Impacts of the ShakeOut Scenario

Transcription

1 CREATE Research Archive Published Articles & Papers Economic Impacts of the ShakeOut Scenario Adam Z. Rose University of Southern California, Dan Wei University of Southern California Anne Wein Follow this and additional works at: Part of the Econometrics Commons, and the Other Economics Commons Recommended Citation Rose, Adam Z.; Wei, Dan; and Wein, Anne, "Economic Impacts of the ShakeOut Scenario" (2011). Published Articles & Papers. Paper This Article is brought to you for free and open access by CREATE Research Archive. It has been accepted for inclusion in Published Articles & Papers by an authorized administrator of CREATE Research Archive. For more information, please contact

2 Economic Impacts of the ShakeOut Scenario Adam Rose, a) M.EERI, Dan Wei, a) and Anne Wein, b) M.EERI For the ShakeOut Earthquake Scenario, we estimate $68 billion in direct and indirect business interruption (BI) and $11 billion in related costs in addition to the $113 billion in property damage in an eight-county Southern California region. The modeled conduits of shock to the economy are property damage and lifeline service outages that affect the economy s ability to produce. Property damage from fire is 50% greater than property damage from shaking because fire is more devastating. BI from water service disruption and fire each represent around one-third of total BI losses because of the long duration of service outage or long restoration and reconstruction periods. Total BI losses are 4.3% of annual gross output in the affected region, an impact far larger than most conventional economic recessions. These losses are still much lower than they potentially could be due to the resilience of the economy. [DOI: / ] INTRODUCTION A major earthquake on the southern San Andreas fault, such as the ShakeOut Scenario (Jones et al. 2008), would have devastating economic consequences for the eight-county region comprising Southern California. 1 Building on estimates of property damage of $113 billion and some estimates of direct business interruption by other members of the research team (see other papers in this volume), we estimate the total business interruption impacts at $68 billion and related costs at nearly $11 billion. This could make the ShakeOut Earthquake Scenario the costliest disaster in U.S. history. Current estimates of Hurricane Katrina are $90 billion in property damage and about $50 billion in business interruption (Swiss Re 2008). The economic impacts of the September 11 attacks on the World Trade Center have been estimated at $25 billion in property damage and $110 billion in business interruption (Grossi 2009, Rose et al. 2009). Many background considerations, disaster characteristics, individual behavioral responses, and workings of the macroeconomy affect economic consequences. One important factor is resilience, or inherent and adaptive actions that mute the losses by using resources more efficiently and recovering more quickly (e.g., Rose 2007, Wein and Rose 2011). Both 9=11 and Katrina were characterized by several forms of resilience, such as business relocation and substitution of imported goods, respectively, that reduced the losses to levels lower than might otherwise have occurred. This is likely to be the case for a major earthquake as well, and was factored into our estimation. a) University of Southern California, 310 RTH, 3710 McClintock Ave, Los Angeles, CA b) U.S. Geological Survey, MS 531, 345 Middlefield Rd., Menlo Park, CA The eight-county region consists of the entirety of Imperial, Kern, Los Angeles, Orange, Riverside, San Bernardino, San Diego, and Ventura counties. 539 Earthquake Spectra, Volume 27, No. 2, pages , May 2011; VC 2011, Earthquake Engineering Research Institute

3 540 A. Z. ROSE, D. WEI, AND A. WEIN This paper explains the estimated economic losses of the ShakeOut Earthquake Scenario. We begin by differentiating the various loss categories and identifying the conduits through which these losses occurred. We then provide an overview of the disaster loss (economic consequence) estimation framework, including a discussion of methods used, assumptions, and sources of data. This includes details of major methodological refinements, including consideration of resilience. We then present and interpret the results. The paper concludes with a discussion of policy implications derived from the analysis. BASIC CONSIDERATIONS DISASTER LOSS ESTIMATION For many years, disaster loss estimation focused on property damage to structures. All other types of impacts (economic, sociological, psychological, etc.) were thrown into a grab bag category termed indirect or secondary losses. By the mid-1990s, there was a growing appreciation of the role of business interruption (BI) losses, which refer to the reduction in the flow of goods and services produced by property (capital stock). This stock=flow distinction is a basic concept in economics and there are direct and indirect versions of both categories. Direct property damage relates to the effects of natural phenomena, such as fault rupture, ground shaking, ground failure, landslides, tsunami, etc., while collateral, or indirect, property damage is exemplified by ancillary fire caused by ruptured pipelines, frayed electrical wires, etc., and exacerbated by loss of water services. Direct BI refers to the immediate reduction or cessation of economic production in a damaged factory or in a factory cut off from at least one of its utility lifelines. Indirect BI (referred to as contingent BI by the insurance industry) stems from the ripple, or multiplier, effects associated with the supply chain or customer chain of the directly affected business. The reader is referred to Rose (2004) for an exposition of these concepts and other related ones below, and to European Union (2003), MMC (2005), National Research Council (2005) and Rose et al. (2007) for examples of their application. An important consideration to emphasize is that nearly all direct property damage takes place at a given point in time (during the ground shaking), and that ancillary (or indirect) property damage takes place during a fairly short time span. BI, on the other hand, being a flow variable, is time-dependent. It begins when the ground shaking starts and continues until the built environment is repaired and reconstructed to some desired or feasible level (not necessarily pre-disaster status) and a healthy business environment is restored. As such, BI is complicated because it is highly influenced by the choices of private and public decision makers about the pattern of recovery, including repair and reconstruction. As in the ShakeOut Scenario, the size of BI can rival that of property damage. More recently, the loss estimation framework has been expanded in three ways, and the term economic consequence analysis is being used to distinguish this breadth (Rose 2009). First, more sophisticated analyses of the ordinary workings and complexities of the macroeconomy are being taken into account. These include not only the ordinary quantity (material) interdependencies through supply chains, but workings of markets and prices, financial variables, and government policy (e.g., Dixon et al. 2010, Rose et al. 2009). Second, is the incorporation of the loss reduction strategy of resilience, in both static and dynamic forms. We define static economic resilience as the ability of an entity or system to

4 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 541 maintain function (e.g., continue producing) when shocked by the types of disruptions accounted for above (see also Rose 2007). It is thus aligned with the fundamental economic problem efficient allocation of resources, which is exacerbated in the context of disasters. This aspect is interpreted as static because it can be attained without repair and reconstruction activities, which affect not only the current level of economic activity but also its future time path. Another key feature of static economic resilience is that it is primarily a demand-side phenomenon involving users of inputs (customers) rather than producers (suppliers). This is in contrast to supply-side considerations, which definitely require the repair or reconstruction of critical inputs. A more general definition of dynamic resilience is the speed at which an entity or system recovers from a severe shock to achieve a desired state. This also subsumes the concept of mathematical or system stability because it implies the system is able to bounce back. This version of resilience is relatively more complex because it involves a long-term investment problem associated with repair and reconstruction. The third major consideration is extended linkages. The first of these refers to systems linkages, such as cascading infrastructure failures. The second is behavioral linkages, which refer to considerations like the effect of recent disasters on risk attitudes (Burns and Slovic 2007). A good example is the fact that 85% of the BI loss following 9=11 stemmed from the nearly two-year decline in air travel and related tourism because of heightened fear of flying (Rose et al. 2009). Note also that this category has associated indirect effects. Thus it can increase ordinary BI losses by one or two orders of magnitude (see also Gordon et al. 2007). In the analysis below, we take into account these various considerations to the extent possible within project limitations. The stock=flow and direct=indirect loss distinctions were made. We included some additional elements of the workings of the macroeconomy, such as the effect on business customers, in addition to the standard emphasis on suppliers. We included a major source of resilience in the aftermath of disasters. We factored in some cascading infrastructure failures. However, we were not able to factor in behavioral linkages. CONDUITS OF ECONOMIC SHOCKS In this paper we analyze the following conduits of shocks to the economic system that include damage to the built environment or interruption of lifeline services: 2 Direct building damage: short-period ground motion (affecting ordinary buildings) Direct building damage: long-period ground motion (affecting high-rise buildings) Indirect building damage: fire following earthquake Direct lifeline service outages for: Electric power Natural gas Water Transportation 2 Actual damage is not necessary in all cases to cause economic loss. Evacuation prior to disaster can cause even greater BI losses than a small version of the event itself. Also, some buildings can be closed for business because of their proximity to damaged buildings. Some infrastructure services may be shut down as a precautionary measure as well.

5 542 A. Z. ROSE, D. WEI, AND A. WEIN An important additional consideration is the need to adjust for double-counting of the conduits of losses. That is, a factory may suffer from a collapsed roof and loss of electricity simultaneously, either one being enough to cause it to shut down business operations. Our analysis does adjust for possible double-counting. Primacy is given to losses from direct building damage. Finally, we note that our results could be presented in terms of several economic impact indicators. We present them in terms of property damage (loss of asset values). We also present them in terms of two types of flow variables relating to BI. The first is gross output, or sales revenue (which equals the cost of all inputs plus a profit term). The second is value added, a net measure that corresponds only to the cost of primary factors of production (labor, capital, and natural resources, and excludes the cost of intermediate, or processed goods). At the regional level, it is the counterpart of Gross National Products (GNP). 3 METHODOLOGY The estimation of the economic impacts of the ShakeOut scenario involves several formal and several ad hoc methods. The estimates of direct impacts made use of a loss estimation model, a highway system damage model, and various calculations for tall buildings. Indirect property damages were estimated by an urban fire (following earthquake) model, and indirect business interruption impacts were estimated with an Input-Output (I-O) model. Moreover, components of this model were used to estimate direct impacts of electricity, gas, and water utility outages and port on-site operation interruption, as well as disruption of import and export through the port. We begin by explaining the I-O model and its application. In the next section the methodological details for each source of business interruption loss are summarized. Many of the cases involve standard application of I-O analysis, but in several cases some ingenuity was required. In the third section we explain how economic resilience is factored into the calculations. We conclude the explanation of the methodology with a discussion of how we adjusted for multiple sources of disruption. INPUT-OUTPUT ANALYSIS I-O analysis, developed by Nobel laureate Wassily Leontief, is the most widely used tool of regional impact analysis in the United States and throughout the world. Moreover, it has been used extensively to analyze the economic impacts of earthquakes and other natural hazards (e.g., ATC 1991, Shinozuka et al. 1998, Rose and Lim 2002, and Gordon et al. 2007). It is especially adept at estimating ripple, or multiplier, effects. I-O can be defined as a static, linear model of all purchases and sales between sectors of an economy, based on the technological relationships of production. Essentially, this is a detailed, comprehensive, double-entry bookkeeping record of all production activity. Practically every country in the world has constructed an input-output table, usually through an exhaustive census or at least an extensive survey, and there is a rich literature on ways to use non-survey data-reduction, or downscaling methods to generate tables for political jurisdictions at various subnational levels. 3 The term Gross here refers to the fact that depreciation (i.e., wear-and-tear or obsolescence of fixed capital assets) is included, although intermediate goods are not included.

6 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 543 I-O analysis provides more than a modeling capability. The basic I-O table of common transactions on which the model is based serves as a valuable framework for organizing an extensive amount of data on a regional economy. The data and analytical tools, such as impact multipliers, derived from it provide insight into the structure, interdependence, and vulnerability of the regional economy (e.g., Miller and Blair 2009). In an I-O analysis, it is important to distinguish two types of second-order effects. The first is indirect effects, which represent the interaction between producing sectors. The second is induced effects, which represent the interaction between households and producing sectors; production generates income paid to households, who in turn spend a major portion of this income on produced goods and services, thereby generating additional multiplier effects. For our model, we accessed the currently most widely used source of regional I-O tables, the Impact Analysis for Planning (IMPLAN) System (MIG 2006). This source consists of three components: 1) a county-level database, 2) a set of algorithms capable of generating I-O tables for any county or county group, and 3) a computational capability for calculating multipliers and performing impact analyses. The IMPLAN sectoring scheme is currently based on the North American Industrial Classification System (NAICS), and the version of the I-O model we used includes more than 500 sectors. We aggregated the sectors to 26 in number because we did not have earthquake-related damage data that would warrant a finer delineation of sectors. Also, using and presenting hundreds of sectors can prove unwieldy. Note that our sectoral classification provides a high level of detail for infrastructure, such as electricity, gas, water, and transportation. It also includes owneroccupied dwellings to incorporate the productive services of homes (equivalent to their rental value) in our estimates. I-O has been used successfully in conjunction with HAZUS, FEMA s hazard loss estimation software (e.g., Rose et al. 2007, Gordon et al. 2007, FEMA 2008). In fact, the Indirect Economic Loss Module (IELM) of HAZUS is based on an I-O methodology. We do not use the HAZUS IELM for several reasons. First, we were able to construct a model at a finer level of sectoral detail than is available in HAZUS. Second, we concluded the IELM involves some assumptions regarding interregional trade that would exaggerate the ability of the economy to adjust to the earthquake and would thus underestimate the impacts. Third, the IELM is not capable of estimating the full set of impacts from infrastructure damage, which constitutes the majority of BI losses from the ShakeOut Scenario. Fourth, we invoke several refinements for resilience that cannot readily be incorporated into the IELM. In particular, the resilience strategy of production rescheduling refers to the ability of businesses to recapture lost production by working overtime or extra shifts once their operational capability is restored and their critical inputs and employees are available. Another source of resilience is utility lifeline importance, which refers to dependency of businesses on electricity, water, gas, and communications. We were not, however, able to incorporate other types of resilience, such as input substitution, conservation, and pricing. These are quite cumbersome to model in an I-O context, but most of these other resilience actions are very small in comparison to the two that were actually modeled (Tierney 1997, Rose and Lim 2002). 4 For both the IELM and our model, I-O analyses are limited by inherent 4 This conclusion is based on survey work by Tierney (1997; see also Rose and Lim 2002). Draconian conservation has been found to apply for disasters with short recovery periods, but not for long ones.

7 544 A. Z. ROSE, D. WEI, AND A. WEIN linearity and lack of behavioral content and reflection of the operation of prices and markets. The inherent rigidity of I-O means that it is likely to under-estimate various flexibilities associated with resilience. Thus, our estimates should be considered an upper bound. At the same time, we can emphasize that they are likely to be a reasonable approximation because we have factored in two of the major sources of resilience. METHODOLOGICAL DETAILS FOR INDIVIDUAL LOSS CATEGORIES In addition to the I-O table, other data are critical for evaluating economic impacts and resilience associated with disasters. These include inventory data on the built environment (factories, residences, infrastructure) and the natural environment. Also needed is a set of damage functions that relate changes in underlying conditions to property damage and loss of function. One such source is FEMA s Hazards United States Multi-Hazard (HAZUS- MH) System (FEMA 2008). This is a large expert system that contains census tract data on the built environment, a set of damage functions, and a GIS capability. Physical damage and business interruption are translated into direct dollar values of building replacement costs and business downtime costs, respectively. Estimation of the conduits of business interruption was as follows: 1. Ordinary buildings. Ordinary buildings as used here pertain to buildings other than high-rises, which were examined separately. The building damage estimates were calculated using HAZUS (see Jones et al. 2008, Graf and Seligson 2011). The flow of goods and services emanating from this productive capital stock (essentially equivalent to BI losses) represent the direct output loss estimates, where output refers to gross output in our I-O Model, equivalent to gross sales revenue. It subsumes other flow categories such as wage, capital-related income, and rent losses calculated by HAZUS as well. HAZUS also calculates relocation costs, which are about $2.8 billion and for which we did not compute indirect and induced effects. We converted HAZUS gross output (GO) estimates into value added (VA) estimates as the bottom line measure of all of our economic impacts by multiplying each sectoral GO estimate by the ratio of VA to GO in that sector. To transform the estimates of direct GO and VA into total GO and VA, we use standard I-O computations. This involves first converting the gross output estimates into final demand, which eliminates a small aspect of double counting of the intermediate goods in each sector. The final demand vector is then pre-multiplied by the I-O inverse matrix, or the matrix of sectoral multipliers. Each element of the inverse matrix represents the total gross output impact on one sector for every dollar change in final demand of another sector. 2. High-rise buildings. The high-rise building study (see Jones et al. 2008, Krishnan and Muto 2011) posited a total of 15 collapsed and red-tagged buildings in four counties. Using input from Porter (2008), we first identified the HAZUS occupancy class to which these buildings belong, and functions of these high-rise buildings. Second, in the absence of HAZUS output losses for high-rise buildings, we based our direct gross output losses on an adaptation of ordinary building loss estimates in terms of square footage estimates for highrises. We assumed the annual gross output (sales revenue) from the economic activities in these buildings is $545.5 per square foot and computed the annual direct output loss by multiplying the total square feet of the damaged buildings by the unit space annual output.

8 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 545 Third, we calculated the indirect and induced effects, and thus the total effects, using the I-O model in the same manner as for ordinary buildings. The replacement of a collapsed or red-tagged building usually takes a long time, ranging from 18 months to three years. 5 Also, the replacement for the nongovernment buildings is not automatic. In the analysis, we assumed that the average time of restoration is 27 months. 3. Fire. First, we obtained the estimates of property damage from the fire following earthquake study (see Scawthorn 2011) with Los Angeles and Orange Counties being by far the most affected. Then, we allocated the burned square footage to HAZUS occupancy classes. The allocation is a function of total square footage burned, number of fires in each county, percent allocated to building uses in the region, and the square footage of each occupancy class in each county (see Chapter 7, Jones et al. 2008). A square foot of fire damage corresponds to a square foot of the HAZUS Complete Damage State (Scawthorn 2008). Therefore, the recovery time of a burned building, or the duration of production disruption for the affected sector, is assumed to be the same as the time to replace a completely damaged building in HAZUS. For different occupancy classes, the recovery time for a completely damaged building ranges from 120 days to 960 days (FEMA 2003). Some sectors in our 26-sectoring scheme cover more than one HAZUS occupancy class. In such cases, we used the average of the recovery time for the relevant occupancy classes. For the industrial and commercial sectors, we calculated the direct output losses from fire in a similar way as for the high-rise buildings, using the same annual output per square foot estimate of $545.5=sq. ft. For residential and education sectors, we adopted a different strategy. For single family homes and schools, we divided the total output from the I-O table by the total square feet for each of them to get a dollar per square foot estimate for different counties. We divided the single-family dollar per square foot estimate in half for the multi-family dwellings, as a crude approximation for this type of dwelling. Again, the estimation of indirect and induced effects was the same as for ordinary buildings. 4. Electric power. Electricity and other infrastructure require a more complicated set of impact analyses. First, there are multiplier effects from the reduced services from the infrastructure provider. Second, there are the direct and multiplier effects from disruption to customers. The first effect requires only an ordinary multiplier analysis as described for the previous loss categories (what we term a demand-side analysis). The second requires the identification of direct losses to infrastructure customers first and then the calculation of multiplier effects, using ordinary multiplier computations (what we term a supply-side analysis). Direct customer economic output losses were ascertained through a methodology developed in the course of another study by the senior author as a supplement to HAZUS, or HAZUS Patch (MMC 2005, Appendix F; Rose et al. 2007). 5 However, some very large and expensive buildings may take longer than 3 years to rebuild. Also, many of the delays are likely related to uninsured buildings, in which case there are fewer financial resources for repair and reconstruction.

9 546 A. Z. ROSE, D. WEI, AND A. WEIN Most of the infrastructure computations involved service disruptions over time. In fact, the percentage of customers affected by the outages is not constant but decrease over time as services are restored. The power restoration pattern (percentage of power services recovered in a series of restoration periods) differed by county group (see Chapter 7, Jones et al. 2008), so the BI impacts were simulated separately for each county group. 5. Water. The estimation of BI losses stemming from disruption of the water system was similar to that of the power system. In this case, we received two types of data to calculate disruption of the water systems by county. An exposure analysis of two-digit NAICS sector employment distribution across ground-shaking intensity zones by county was used to approximate the percentage of sector production that is located in different ground-shaking affected regions (see Sherrouse et al. 2008). The second type of data described the percentage water supply outage and restoration time by ground-shaking intensity zone. The disruption rate and restoration pattern differed by ground-shaking intensity zone: there is no disruption at all in Instrumental Intensity V and below areas; 20% of customers in Instrumental Intensity VI & VII areas are without water for two weeks; 50% of customers in Instrumental Intensity VIII areas are without water for two weeks and the water outage will last for another 2.5 months for 20% of the customers; and 100% of customers in Instrumental Intensity IX & X areas are without water for six months. Most sectors of San Diego County did not experience any water service disruption because they are located in areas where the ground-shaking intensity is below Instrumental Intensity V. For those counties with area within the Instrumental Intensity VIII or higher ground-shaking zones, more businesses would suffer from longer water service disruption up to six months. For example, in San Bernardino County, 23% and 11% of the total businesses are located in Instrumental Intensity IX & X ground-shaking areas, respectively. Therefore, 34% of the total businesses of the San Bernardino County would experience 6-month water disruption. Sector activity is not evenly distributed across the ground shaking intensity zones and Figure 1 depicts the percentage of sector employees working in zones of instrumental intensity VIII to X for the eight-county region. The analysis required a modification because water systems are predicted to be more extremely damaged than power systems. Type II multiplier for the eight-county area are around 2.0, meaning that a 20% reduction in direct gross output in one sector translates into a doubling of that impact for the region as a whole. However, this is problematic when the direct reduction is high for any or group of sectors. Accordingly, we capped the total gross output loss at 100 percent. 6. Natural gas. Disruption of gas service by county is recorded in Chapter 7, Jones et al The methodology was the same as the power system estimation. In the direct output loss calculation, for the residential sectors, we multiplied the outage ratios by the percent sold to households (0.95). For all the other sectors, we multiplied the outage ratios by the percent sold to business (0.05). In Imperial, Kern, San Diego, and Ventura counties 100% of the gas service was restored after three weeks. In Los Angeles, Orange, Riverside, and San Bernardino counties, 92% to 97% of the gas service was restored after three weeks and 100% was restored after two months. Again, because the restoration pattern differed by county, each set of county BI impacts were simulated

10 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 547 Figure 1. Sector sensitivity: Percent sector employees in Instrumental Intensity VIII-X zones based on 2006 fourth quarter economic data. Source: California Economic Development Department (2008). separately. Estimates of indirect and induced effects used the same methodology as for electric power. 7. Transportation. Direct highway damage estimates, truck and passenger traffic impact as well as the cost of traffic delays, were obtained from Werner et al. (2008) and Sungbin Cho (see Chapter 7, Jones et al. 2008). However, we were not able to compute indirect and induced effects for the delays because of the difficulty of assigning delays to individual sectors. The original data were provided as final demand reductions across the 13 Southern California Association of Government (SCAG) industrial sectors for 6 counties that were significantly affected. For each of the 6 counties, the final demand reductions data were derived for three restoration periods: the first three months, the second three months, and the final 40 days. We allocated the final demand reductions by restoration period to the 26 sectors using sectoral total output as weights. Again, we applied our HAZUS-Patch methodology. Estimates of indirect and induced effects used the same methodology as previously described. 8. Ports. Estimation of direct BI loss from disruption of ports involved two steps: the losses due to the reduced supply of imported production inputs and the losses due to the final demand reduction in foreign exports. For both steps, we assumed that the entire disruption period was seven weeks (see Chapter 7, Jones et al. 2008). For the losses pertaining to the imported production inputs, we first computed the percent of inputs in each sector that are foreign imports using IMPLAN data and the following formula: % inputs from foreign imports ¼ foreign imported inputs=(regionally produced inputs þ domestic imported

11 548 A. Z. ROSE, D. WEI, AND A. WEIN inputs þ foreign imported inputs). The percentage of foreign imports that come into California by ships is 76%, so we used 80% for the Southern California because of its proximity to the port. Therefore, if a sector receives 10% of its inputs from foreign imports, then because of the port shutdowns, the reduction in production inputs for this sector would be 8%. Given the linear relationship between the production input and output of the I-O model, the direct gross output loss for this sector is 8%. We also assumed that in the first three days, there are sufficient inventories (a form of resilience) of foreign imports, so there is no output reduction in any sector. The total volume of exports through the ports each year is about $36 billion, but only about $7.2 billion is produced in the eight-county region. Based on this annual total volume of exports, we computed the total export demand loss during the ports disruption period (35 days of predicted downtime). This total loss in foreign export demand is then distributed among the seven goods producing sectors using the IMPLAN data in the foreign export demand column. We computed the direct gross output loss of the Water Transportation sector differently from the other infrastructure computations. The direct loss in this sector was calculated by excluding the total services to the household sectors from the sectoral total output because households make up such a small portion of its output. Estimation of indirect and induced effects for both the foreign import input disruption and the foreign export demand reduction followed the standard set of computations for other infrastructure sectors. RESILIENCE For the most part, this study, only addresses aspects of static resilience because of the limitations of data and because dynamic resilience, especially for infrastructure, is so strongly dependent upon complex public and private decisions regarding the timing of repair and reconstruction, and hence highly variable. Moreover, only two of the static resilience options were incorporated, albeit the ones that have been found to have the greatest potential for reducing BI losses (Rose et al. 2007). The first of these is production recapture or rescheduling, the ability of firms to work overtime or extra shifts after they have repaired or replaced the necessary equipment and their employees and critical inputs become accessible, that is, once loss of function has been eliminated. In an earthquake context, this is rather straightforward for the case of building damage. For infrastructure, it is more complicated. Just because electricity service has been restored doesn t mean that businesses can immediately turn on the assembly line; they must repair the necessary plant and equipment first (though this need not be 100% restoration to be fully operational). HAZUS includes an adjustment for this consideration, referred to as the Building Service Interruption Time Multiplier. Production rescheduling was first formally incorporated into HAZUS loss estimation by the senior author and Stephanie Chang through the inclusion of production recapture factors (RFs). The scalars represent the percentage of direct gross output losses that can be recovered at a later date. The original HAZUS RFs range from.30 to.99. Manufacturing enterprises that produce nonperishable commodities are at the high end, while sectors producing perishables (e.g., agricultural) or non-essential services (e.g., entertainment) are at the lower end of the scale. The original RFs came with a caveat that they were applicable

12 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 549 only for three months, and that they would dissipate thereafter. This refers to the fact that customers and suppliers will grow impatient as their orders go unfilled. Accordingly, we adjusted the HAZUS RFs downward by a linear decay rate of 25% for every three-month period during the first year, with the last reduction lasting through the second year. Each RF is effectively zero thereafter. Loss of function duration is a major data input to apply to the modification of the RFs. This was provided for ordinary buildings as part of the original HAZUS computations. The rebuilding time for tall buildings and for buildings damaged by fire damage was essentially larger than two-years, so the RFs were zero for these categories. Duration data for infrastructure is more straightforward it is simply the length of the outage, and, as noted earlier, the percentage of customers affected is not constant. For the port simulation cases with recapture adjustment, we used a recapture factor of 85% (to account for ship diversions, cancelled shipments, and declined bookings) for all the sectors except for Agriculture. For the Agriculture sector, we assumed recapture factors of 50% for the first three days, 25% for the fourth day to the second week, and 10% for the second to seventh week, a rapid decay in the ability to recapture because of the perishable nature of agricultural commodities. Overall, production rescheduling resulted in sizeable reductions in direct BI losses for various categories. This ranged from a high of 85% for electricity outages to a low of 29% for fire damage. Again, it is not the building=infrastructure type itself that is the dominating factor, but the length of the disruption associated with it. The second type of resilience modeled was infrastructure importance. The term stems from ATC-25 (1991), which convened a panel of experts to advance hazard loss estimation. One of the contributions was to identify the percentage of a sector s business operations that is not dependent on a specific infrastructure type. Thus, even if there is a lifeline outage, a portion of the sector can keep operating. Examples are headquarters offices being less dependent than production lines in general, and some sectors being less dependent than others on lifeline services (e.g., the relatively low dependence of agriculture on electricity and natural gas). Typically, businesses are most dependent upon electricity, then water, then natural gas. Like production rescheduling, this type of resilience also dissipates over time, though in a less dramatic manner. For example, if activities of headquarters or maintenance facilities are disrupted, other business functions may still be able to continue, but eventually inoperable headquarters and maintenance operations will disable the other functions of the enterprise. Unfortunately, no data were available to make any reasonable adjustments in the importance factors to reflect a decline in this type of resilience. ADJUSTMENT FOR MULTIPLE SOURCES OF BUSINESS INTERRUPTION Many businesses and households will suffer shocks from many sources. They may simultaneously incur building damage and loss of one or more lifeline services. Thus, each of our estimates when totaled may double-count some impacts the same business establishment cannot be shut down more than once in any given period. We adjusted for these multiple causes of failure, by first noting time periods for which these would take place, typically short periods of time. We then took the ratio of time periods for various sources of shocks and subtracted the ratio from the total business interruption impacts. From an

13 550 A. Z. ROSE, D. WEI, AND A. WEIN examination of the issues surrounding this problem in the literature, (e.g., Chang 2008) however, we assumed that half of the cases when two or more shocks occurred simultaneously involved redundancies, or double-counting. Thus, the probability that the same business in a given county was actually hit by the two shocks modeled for the county was 0.5. SHAKEOUT SCENARIO ECONOMIC IMPACT The major results of our study are presented in Table 1. The first partition of the table pertains to property damage. The largest category of direct damage is that of Ordinary Buildings, at $32.7 billion. Related Content Damage is $10.6 billion, or about one-third the building damage. Damage to High-Rise Buildings is rather small in comparison, at $2.2 billion. Again Content Damage is about one-third this size. The largest total property damage category, however, is Fire Damage to Buildings, at $40 billion. This damage is expected to Table 1. Economic impacts (with recapture; in billions of 2008 dollars) Indicator Direct Impacts Total Impacts Ordinary Building Damage $32.7 $32.7 Ordinary Building Content Damage High-Rise Building Damage High-Rise Content Damage Fire Damage Fire-Related Content Damage Highway & Pipeline Damage Subtotal Property (stock) Damage BI from Ordinary Buildings BI from High-Rise Buildings BI from Fire BI from Power BI from Water BI from Gas BI from Transportation BI from Ports Subtotal BI (Flow) Loss Relocation Costs Traffic Delay Costs Subtotal Additional Costs Total a $96.9 $190.9 a Does not include non-quantifiable losses.

14 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 551 be widespread, because of close proximity of buildings and extensive stock of wood frame structures. Content Damage for this category is relativity high at $25 billion, or 62.5% of building damage due to fire. The main reason for the relatively higher ratio than that of ordinary buildings is the greater likelihood of total destruction in the case of fire. Damage to Highways and Pipelines is rather small at $1.5 billion. However, the ramifications for business interruption are huge, especially for the damage to the pipeline transportation of water. Total property damage is estimated at $112.7 billion. The largest category of direct BI is for the interruption of Water Services, at $13.8 billion. Multiplier effects averaging at 1.77 translate this into total impacts of $24.2 billion, including $10.3 billion of indirect and induced effects. As is the case with most forms of BI, these impacts are primarily due to the extent of property damage (not necessarily the cost) and the time it takes to recover. In the case of water systems, many areas could possibly be without ordinary piped water delivery for as much as six months. Table 2 presents the sectoral output losses for water service disruption alone in the eight-county region. Direct BI for Fire Damage is the next largest category, estimated at $12.8 billion, with multiplier effects of 1.75 accumulating a total BI impact of $22.4. Direct BI from Electricity outages is at $4.4 billion, followed closely by direct BI for damage to Ordinary Buildings, at $4.3 billion. However, because the sectors most affected by Ordinary Building damage have a higher average multiplier than those that suffer from Electricity outages, total losses are larger for the latter category. BI losses from Natural Gas disruptions are relatively low because businesses are less dependent on this lifeline. BI from Ports is rather low because of the distance of these facilities from the epicenter of the earthquake and the ability of the port to eventually clear the backlog. 6 We also assumed that direct and indirect port customers have sufficient inventories of goods to negate any adverse effect during the first three days of the disruption. BI from Transportation does not translate into outright reductions in business activity but is mostly captured in the $4.3 billion Traffic Delay cost. Finally, Business Relocation costs are estimated using HAZUS at $6.4 billion, and likely reflects a bias that firms prefer repairing and rebuilding rather than relocating. However, while warranting greater scrutiny, it is unlikely that the opposite assumption would have had much effect on the overall results. Total BI losses are thus $67.5 billion, or 40% lower than property damage. Note that all of the BI losses are calculated after factoring in the effects of recapture of lost production by working overtime or extra shifts after the earthquake. Total BI losses would have exceeded more than $200 billion were it not for this source of resilience. 7 Adjustments for possible double-counting of multiple failure modes would reduce the BI losses from $67.5 billion to $61.3 billion. The result of this adjustment for double- 6 Still, we have probably underestimated the port impacts due to the complexities of modeling the importance of this infrastructure category to the Southern California economy. However, we may also have underestimated resilience because a subsequent study by Northcom suggested more ships could be diverted to other ports than assumed. 7 The results presented in Table 1 corrected two errors in the losses reported in Jones (2008). One is based on a correction of an omission in HAZUS that has only recently been identified. This caused us to increase our estimate of relocation costs from $0.1billion to $6.4 billion. The second was due to a prior misunderstanding between research team members that caused us to recalculate direct and indirect losses from water supply disruptions and change our total BI estimate attributable to this source from $53 billion to $24 billion.

15 552 A. Z. ROSE, D. WEI, AND A. WEIN Table 2. Total output losses by sector for water disruption alone (in millions of 2008 dollars) Without Recapture With Recapture Sector Total Output Loss Percentage Annual Total Output Loss Total Output Loss Percentage Annual Total Output Loss 1 Agriculture % % 2 Construction 3, % % 3 Food, Drugs & Chemicals 13, % 2, % 4 Mining & Metals=Minerals 3, % % Processing & Mft. 5 High Technology 1, % % 6 Other Heavy Industry 9, % 1, % 7 Other Light Industry 7, % 1, % 8 Air Transportation % % 9 Rail Transportation % % 10 Water Transportation % % 11 Highway & Light Rail 1, % % Transportation 12 Electric Utilities 1, % % 13 Gas Utilities 1, % % 14 Water Utilities % % 15 Wholesale Trade 4, % 1, % 16 Retail Trade 4, % 1, % 17 Banks & Financial Institutions 1, % % 18 Professional & Technical Services 11, % 2, % 19 Education Services 1, % % 20 Health Services 4, % 1, % 21 Entertainment & Recreation 7, % 2, % 22 Hotels % % 23 Other Services 2, % % 24 Gov t & Non-NAICS 2, % % 25 Real Estate 2, % 1, % 26 Owner-occupied dwellings* 4, % 2, % Total 93, % 24, % counting was relatively minor a 9.1% decrease in the overall BI estimate for our overlap factor of 0.5. This is mostly due to the fact that the dominating sources of impacts water service disruptions and fire damage were the only source of shock for long periods of time, typically more than four months. Moreover, these two sources themselves had minimal overlap, since the areas affected by each were distinct (LA and Orange Counties for fire damage, and San Bernardino and Riverside Counties for water disruption). We also computed an upper bound adjustment to see how sensitive the results might be to double-counting. However, even if we used an overlap factor of 0.9 rather than 0.5, the downward adjustment for double-counting would be only $56.4 billion, or 16.4%.

16 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 553 In absolute terms, the sectors most affected are Manufacturing (especially Food & Chemicals), Professional, Scientific and Technical Services, Entertainment and Recreation, Wholesale and Retail Trade, Health Services, and Owner-Occupied Dwellings. These sector losses can largely be explained by five key factors: the size of sector (in terms of output), the extent to which other sectors are dependent on it directly and indirectly (multiplier impacts on the sector), the extent the sector s buildings suffer fire damage, their reliance on water, and=or the sector recapture factor. For example, the output from Professional, Scientific, and Technical services and Wholesale and Retail Trade is relatively high. Fire damage has a bigger effect on Professional and Technical Services, Entertainment and Recreation (through restaurants) and Residential Building occupancies. The importance of water is particularly high for Food and Chemicals and for Entertainment and Recreation. In addition, the recapture factor is lower for Entertainment and Recreation and Health Services and zero for Owner-Occupied Buildings. The sectors most sensitive to the earthquake shocks in terms of percent output loss are Hotels, Water Utilities, Owner-Occupied Dwellings, and Mining, Minerals Processing, and Metals Manufacturing. Sector sensitivity is strongly influenced by the sector s reliance on water and the sector recapture factor. Yet another comparison is with the concept of a recession. The basic definition is a drop in GNP over the course of two successive calendar quarters. Nearly all the post- World War II recessions until the 1990s actually exceeded a 2% drop in GNP at some point of their duration. There is growing appreciation that recessions should also be demarcated for regions and not just the U.S. economy as a whole. Accordingly, then the ShakeOut Scenario BI impacts qualify as a regional recession by either definition. Recessions receive a great deal of attention and are the cause of great concern at the national level. There is every reason to express similar concern at the regional level. The fact that the downturn is caused by a natural disaster or a terrorist attack is no less important than if it were caused by ordinary business cycles, financial manipulation, or other historical causes of recessions. POLICY IMPLICATIONS Below we list some of the major economic impacts of the ShakeOut Scenario and their implications for policy: Of the $191 billion of losses, only the building damage and infrastructure damage occur at the time of the earthquake; fire damage ensues for a couple of days and business interruptions occur over weeks and months. Policies should be cognizant of the timing of losses and implications for actions. Mitigation is effective to reduce property damage and ensuing BI, while only resilience is effective thereafter. Damage to buildings and their contents from fire following the earthquake are nearly 50% greater than losses from ground-shaking. BI losses are about two times as great, and indirect losses almost match the direct losses. Containment of these fires is initially determined by response before it becomes part of the recovery process. Mitigation of fire damage in this case takes place before and after the event, and both should be accorded high priority, as well as maximizing fire containment when there are water constraints and other limited resources.

17 554 A. Z. ROSE, D. WEI, AND A. WEIN The largest category of BI stems from disruption of water services. Resilience measures are more limited for this key resource because it has few substitutes and because of the massive need for it. Moreover, the projected lengthy period of disruption diminishes the effectiveness of one of greatest sources of resilience production recapture. Preparing for interim solutions and speeding recovery, repair and reconstruction efforts to restore water service should be a very high priority. Water and power sector output losses are only 1% of the total regional output losses they cause; 99% of water-related losses are experienced by water and power service customers. Lifeline operators thus need incentives to reduce service outages via mitigation or business continuity planning beyond their own narrow business perspective, i.e., to take into account the impacts on the rest of the economy. Multiplier effects of direct BI impacts increase the direct loss total by 76%. Moreover, these effects are likely to be felt in all reaches of the eight-county region, even in areas of very light ground shaking. For example, impacts of the transportation system, including lost and delayed truck and passenger trips and halted goods movement by rail in both directions, will cause congestion at the ports despite little direct damage. Assistance to businesses throughout the region is warranted past the emergency phase, and business continuity planning should be further encouraged. Various types of resilience erode with time. Many guidelines for business continuity or recovery planning focus on Day 30 or 60. More focus on the time aspects of recovery and resilience will assist with planning for worst case or a catastrophic event when the recovery period is on the order of years. CONCLUSION The modeling approach summarized here is capable of estimating not only the apparent direct effects but also the regional indirect economic effects of earthquake damage. The avoidance of these consequences corresponds to economic benefits of hazard mitigation and resilience. The application of these models can help identify the lowest cost strategies for reducing economic losses. Most prior analyses have focused on mitigation, but the newer concept of resilience warrants attention. Some resilience options are relatively low cost (e.g., production rescheduling), some may even be cost-saving (conservation), and most of them need not sit idly in anticipation of an event, but can be marshaled when needed. It is impossible to protect the general population against all natural disasters and terrorist attacks. However, individuals, firms, and government agencies can protect themselves from the negative impacts of business and infrastructure disruptions by implementing various types of resilience (see Wein and Rose 2011). Thus, in benefit-cost analyses of ways to reduce losses from disasters, there is a need to take a holistic view of trade-offs between mitigation and resilient strategies, both of which can significantly result in cost-savings to society as a whole. In the same vein, it is important not to neglect regional economic interdependence effects, including the potential negative effects of the failure of one type of infrastructure upon others. Such interdependencies can significantly raise the stakes at risk. The methodology presented here can provide reasonable estimates of these complex considerations.

18 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 555 The BI estimates were produced for a single plausible outcome of the ShakeOut Earthquake Scenario for the purpose of an emergency response and recovery exercise. The results of the economic impact analysis reinforced the potential benefits of mitigation and resilience strategies for water service providers and fire departments. Also, noteworthy, is the role of this analysis in scenario development. It demanded many inputs of value to the emergency management exercises (e.g., lifeline restoration curves). These would not have been pursued with the same tenacity in the absence of the economic impact analysis. There was not the time or budget to explore uncertainties throughout the development of the scenario from the earthquake rupture to the triggered hazards to the damages and, finally, to the economic impacts. Furthermore, the focus of the economic impact analysis was on analyzing a wide range of conduits for business interruption rather than alternative outcomes for a smaller number of them. Sensitivity analyses would enhance the value of the methods developed and described in this paper. Future economic impact analyses need to expand upon the opportunities for sensitivity analysis including variations in sectoral multipliers, static resilience assumptions (e.g., recapture rates and lifeline importance factors) and, in addition, dynamic resilience alternatives (e.g., pathways to restore infrastructure and=or function). ACKNOWLEDGMENTS We received essential inputs for the conduits of shock from other members of the scenario development team, experts, and stakeholders. Keith Porter produced data and advice at several critical junctures of the research. David Hester and Ben Sherrouse provided spatial analyses of business and goods movement data. Hope Seligson provided direct property damage and direct business interruption estimates for ground-shaking damage. Charles Scawthorn estimated direct damage property damage from fire, and Sungbin Cho provided estimates of direct transportation disruption costs. REFERENCES Applied Technology Council (ATC), Seismic Vulnerability and Impacts of Disruptions of Utility Lifelines in the Cotermininous United States, Report ATC 25, Applied Technology Council, Redwood, CA. Burns, W. J., and Slovic, P., The diffusion of fear: Modeling community response to a terrorist strike, JDMS: The Journal of Defense Modeling and Simulation: Applications, Methodology, Technology 4, available at California Economic Development Department, Special data product request for 2006 fourth-quarter business establishment, employee, and payroll data by ZIP code. Chang, S. E., Pasion, C., Tatebe, K., and Ahmad, R., Linking Lifeline Infrastructure Performance and Community Disaster Resilience: Models and Multi-Stakeholder Processes, MCEER , available at Linking-Lifeline-Infrastructure-Performance-and-Community-Disaster-Resilience-Modelsand-Multi-Stakeholder-Processes-MCEER html. Dixon, P. B., Lee, B., Muehlenbeck, T., Rimmer, M. T., Rose, A., and Verikios, G., An H1N1 epidemic in the U.S.: Analysis using a quarterly CGE model, CoPS=Impact Working Paper No. G202, Monash University, Melbourne, Australia.

19 556 A. Z. ROSE, D. WEI, AND A. WEIN European Union, Proceedings of the Joint NEDEIS and University of Twente Workshop: In Search of a Common Methodology for Damage Estimation, Brussels, Office for Official Publications of the European Communities. Federal Emergency Management Agency (FEMA), Multi-Hazard Loss Estimation Methodology Earthquake Model, HAZUS-MH MR3 Technical Manual, available at (last accessed 23 July 2010). Federal Emergency Management Agency (FEMA), Earthquake Loss Estimation Methodology HAZUS-MH MR3 (HAZUS), National Institute of Building Sciences, Washington, D.C Gordon, P., Richardson, H., Moore II, J., Park, J., and Kim, S., Economic impacts of a terrorist attack on the U.S. commercial aviation system, Risk Analysis 27, Graf, W., and Seligson, H., Damage to wood-framed buildings, Earthquake Spectra 27. Grossi, P., Property damage from the World Trade Center attack, Peace Economics, Peace Science, and Public Policy 15, Article 3. Jones, L. M., Bernknopf, R., Cox, D., Goltz, J., Hudnut, K., Mileti, D., Perry, S., Ponti, D., Porter, K., Reichle, M., Seligson, H., Shoaf, K., Treiman, J., and Wein, A., The ShakeOut Scenario: USGS Open File Report and California Geological Survey Preliminary Report 25, and Sacramento, CA. Krishnan, S., and Muto, M., Hope for the best, prepare for the worst: Response of tall steel buildings to the ShakeOut Scenario, Earthquake Spectra 27, Miller, R., and Blair, P., Input-Output Analysis: Foundations and Extensions (2nd ed.), Cambridge University Press, New York, 782 pp. Minnesota IMPLAN Group (MIG), Impact Analysis for Planning (IMPLAN) System, Stillwater, MN. Multihazard Mitigation Council (MMC), Natural Hazard Mitigation Saves: Independent Study to Assess the Future Benefits of Hazard Mitigation Activities, Study Documentation, Vol. 2, Report to the Federal Emergency Management Agency by the Applied Technology Council, National Institute of Building Sciences, Washington, D.C. National Research Council, Improved Seismic Monitoring Improved Decision-Making: Assessing the Value of Reduced Uncertainty, National Academy Press, Washington, D.C. Porter, K., Personal communication. Rose, A., Economic principles, issues, and research priorities of natural hazard loss Estimation, in Modeling of Spatial Economic Impacts of Natural Hazards, Y. Okuyama and S. Chang (eds.), Springer, Heidelberg, Germany. Rose, A., Economic resilience to natural and man-made disasters: Multidisciplinary origins and contextual dimensions, Environmental Hazards 7, Rose, A., A framework for analyzing and estimating the total economic impacts of a terrorist attack and natural disaster, Journal of Homeland Security and Emergency Management 6, Article 6. Rose, A., and Lim, D., Business interruption losses from natural hazards: Conceptual and methodology issues in the case of the Northridge earthquake, Environmental Hazards: Human and Social Dimensions 4, Rose, A., Oladosu, G., Lee, B., and Beeler-Asay, G., The economic impacts of the 2001 terrorist attacks on the World Trade Center: A computable general equilibrium analysis, Peace Economics, Peace Science, and Public Policy 15, Article 4.

20 ECONOMIC IMPACTS OF THE SHAKEOUT SCENARIO 557 Rose, A., Porter, K., Dash, N., Bouabid, J., Huyck, C., Whitehead, J. C., Shaw, D., Eguchi, R. T., Taylor, C., McLane, T. R., Tobin, L. T., Ganderton, P., Godschalk, D., Kiremidjian, A. S., Tierney, K., West, C. T., Benefit cost analysis of FEMA hazard mitigation grants, Natural Hazards Review 8, Scawthorn, C.R Personal communication. Scawthorn, C. R., Fire Following Earthquake, Earthquake Spectra 27, Sherrouse, B., Hester, D., and Wein, A., Potential Effects of a Scenario Earthquake on the Economy of Southern California: Labor Market Exposure and Sensitivity Analysis to a Magnitude 7.8 Earthquake, U.S. Geological Survey Open File Report, OFR pubs.usgs.gov/of/2008/1211/. This publication is online only. Shinozuka, M., Rose, A., and Eguchi, R., Engineering and Socioeconomic Impacts of Earthquakes: An Analysis of Electricity Lifeline Disruptions in the New Madrid Area, Multidisciplinary Center for Earthquake Engineering Research, Buffalo, NY. Swiss Re, Disaster Losses, 6losses & l ¼ any & search=search & search=search Tierney, K., Impacts of recent disasters on businesses: The 1993 Midwest floods and the 1994 Northridge earthquake, in Economic Consequences of Earthquakes: Preparing for the Unexpected, B. Jones (ed.), National Center for Earthquake Engineering Research, Buffalo, NY. Wein, A., and Rose, A., Economic resilience lessons from the ShakeOut Earthquake Scenario, Earthquake Spectra 27, Werner, S. D., Cho, S., Eguchi, R. T., The ShakeOut Scenario Supplemental Study, Analysis of Risks to Southern California Highway System, U.S. Geological Survey, Menlo Park, CA. (Received 1 January 2010; accepted 10 March 2011)