economic analysis of the consequences of a...
TRANSCRIPT
ECONOMIC ANALYSIS OF THE CONSEQUENCES OF A POTENTIAL
ANTHRAX TERRORIST ATTACK IN THE PACIFIC NORTHWEST:
BUSINESS INCOME LOSSES AND REAL ESTATE PRICE DECLINES
by
Adam Rose (PI)
Noah Dormady
Morgan Bender
Thomas Szelazek
Dan Wei
David Bennett
Final Report to U.S. Department of Homeland Security
Center for Risk and Economic Analysis of Terrorism Events (CREATE)
University of Southern California (USC)
Los Angeles, CA 90089
December 23, 2011
1
CHAPTER 1. OVERVIEW
by
Adam Rose
1. INTRODUCTION
An anthrax attack on a major metropolitan area of the U.S. such as Seattle would have
devastating direct consequences. These would include massive death and injury, contamination of
buildings, and cessation of economic activity in the affected area. Immense sums would need to be
expended on health care and decontamination. The indirect effects would likely be even greater.
Ordinary resources losses stemming from ripple, or multiplier, effects would more than double the
declines in economic output, employment and income. Property value decreases would take place not
only in the affected area, but in the rest of the economy due to the ensuing economic downtrend. Further
economic losses would stem directly and indirectly from fear. Many people would avoid the affected
area and even neighboring areas out of concern for the spread of contamination, thus spatially broadening
the decline in property values. In addition to the short-term effects from the social amplification of risk,
long-run stigma effects could last for many years. A recent study comparing ordinary resource loss and
behavioral effects of a radiological dispersion device (RDD) attack in a major metropolitan area, Los
Angeles, estimated that the latter type of effects could be more than an order of magnitude greater than
the former if stigma effects were to last at least a decade (Giesecke et al., 2011).
The situation could very well lead to the creation of "black holes" in the middle of a major city
attacked by use of an insidious mechanism involving radiological, biological, or chemical agents. This
would obviously be a great psychological, cultural, personal and economic loss to individuals, the region,
and the nation as a whole. While major efforts to prevent such attacks are a high priority, some attention
needs to be placed on how to minimize the short-term and long-term consequences of such events.
Enhancing resilience is one approach. This would involve individuals and group (both private and public
sector) efforts to use remaining resources as efficiently as possible and to hasten the speed of recovery
(including expediting the clean-up process) (Rose, 2009). Individual initiative would help promote
recovery, but the obstacles would be greater in this case than in many others. Just fear of long-lasting
anthrax contamination would lead many people to leave the Seattle area permanently. A precipitous drop
in real estate prices would put many business, apartment, and home owners "under water" in their
mortgages, causing them to consider defaulting and starting fresh in another city or state. Even those
remaining might require incentives to return to or visit the affected area. Workers might require higher
2
wages and investors require a higher rate of return. Resident shoppers and tourists would likely require
price discounts in restaurants and hotels, as well as goods produced in the contaminated area. This would
significantly raise the cost of doing business, which, in turn, would decrease local purchasing power and
make Seattle's exports less competitive. These various factors are what underlie the order of magnitude
increase in negative economic consequences in the LA RDD example.
II. THE NEED FOR INCENTIVES
The increased premiums and discounts working their way through the ordinary market process
provide signals for the efficient allocation of resources, but are still likely to lead to a slow and possibly
ineffectual recovery process. For example, the increased risk of rebuilding or restoring a business in
downtown Seattle may be viewed as so high that the necessary rate of return on investment would be
economically untenable, i.e., no business could earn high enough profits to warrant reopening.
This poses a serious public policy dilemma. Does government enter the picture in order to
provide incentives to re-establish the population economic base of Seattle? Although recovery may fail
the market test, broader societal objectives might warrant an infusion of various forms of subsidies.
These objectives could include national pride, concerns for disaster victims (who have joined the ranks of
severely economically disadvantaged), and even concerns about the drag on the national economy
stemming from an empty Seattle, a large displaced population, and large induced migration. The latter
implies it may be less expensive nationally to subsidize Seattle residents and businesses than to pay
private out-of-pocket or government financed relocation costs and unemployment insurance, or incur
potential costs of severe disruption of families and strains on other regions to absorb the migration.
A large number of incentives could be applied:
• Mortgage payment subsidies
• Low-interest loans
• Decontamination and testing subsidies
• Safe building guarantees/buyouts
• Investor subsidies
• Wage subsidies
• Personal and business income tax credits
• Job retraining
This study will not explicitly examine the impacts of these various alternatives, but will offer key
methods and insights needed to evaluate and refine subsidies in the event of an anthrax attack.
Specifically, we develop methods to estimate the decline in property values and the decline in the return
on investment that would ensue from a major disaster like an anthrax attack. The research in this report
3
adapts conventional economic and finance analysis to the case in point. It complements the report by
John et al. (2011), which primarily uses risk and decision analysis methods, to evaluate how Seattle
residents perceive the post-disaster situation. Both risk and consequence analyses are needed to design
appropriate disaster recovery policies.
III. SUMMARY
This report consists of three substantive chapters. It focuses on two major considerations that are
difficult to assess using risk and decision analysis methods. The methods and results can be used as a
cross-check on other approaches, including surveys.
We simulate the effects of two scenarios:
1. Scenario 1-- Significant Casualty Rate. The DHS Interagency Biological Restoration
Demonstration (IBRD) Scenario projects 54,000 deaths as a result of the anthrax attack. In this
simulation, we use deaths, injuries and direct contamination data as inputs into our models. That is we
simply assume a population decline of 54,000, with no out-migration. The other feature is the closing of
the CBD to economic activity during the decontamination period. This is a very conservative
interpretation of an anthrax attack scenario (AAS) and thus yields a lower-bound estimate.
2. Scenario 2 -- Major Out-Migration. This simulation includes a 75 percent population flight
from Zone 1. The reduction was deduced from survey respondents’ reactions to the IBRD scenario and a
written script of events that take place in the aftermath of the AAS (see Burns et al., 2011). It should be
noted that this migration is not permanent for a portion of that 75 percent population, who are expected to
begin returning after 3 to 4 months. An out-migration of this magnitude would essentially cause the
City's economy and real markets to implode. No model could accurately predict the decline in real estate
values for this extensive of an event, and hence so great a departure from the historical statistical
database. Suffice it to say, however, that a 75 percent population decline would lead to property value
declines of at least that magnitude. We simulate a modified version of this scenario, which allows us to
illustrate the use of the models. This is to assume the temporary exit of 75 percent of just Zone 1
population, which represents slightly over 35 percent of the total population in the 3 zones. This
provides a mid-range estimate of real estate and business rate of return impacts. A simulation of the flight
of 75 percent of the population of Zone 1alone would represent an upper bound.
One way to estimate factors such as wage and investor premia, or shopper and tourist discounts is
through stated preference methods. These typically involve surveys, experiments, or elicitations of
subjects. In many cases the subjects are a cross-section of the population, or even just student volunteers,
rather than a distinct group such as businessmen. In some cases, an expert judgment is not required, such
as in asking a typical respondent how much of a price discount he/she would require to dine at a
4
downtown Seattle restaurant, or even what wage premium would be needed to return to work there.
These are matters of personal choice and/or tastes. However estimation of investor rate of return or the
likely plummet in real estate prices after an anthrax attack require expert judgment. Even if the
respondents are businessmen and real estate agents, the questions are of such a technical nature that a
casual on-the-spot response is unlikely to be very accurate. This is especially the case in an extreme
disaster where no prior experience can be called upon. The alternative is to use models that simulate an
economic or decision process, as in the use of a dynamic income or financial model of an affected
business, or a predictive model, such as a real estate forecasting model.
The next two chapters of this report involve the real estate forecasting model. Chapter 2 presents
the time series of real estate data and the regression analysis methods used. The analysis breaks the
Seattle Metropolitan Area into three zones: Anthrax attack (contamination), fringe, and outlying zones.
This is useful to distinguish the effects of the attack on real estate prices throughout the Seattle area. Key
explanatory variables include area-wide personal income, median household income, employment,
housing inventory, housing purchases/sales, and number of foreclosures. For example, Median
Household Income provides perhaps one of the strongest macro drivers of real estate prices, as it does
throughout individual zones as well. Holding all other variables constant at their mean, a $1,000 increase
in median household income would increase the median sales price of residential property in the mean zip
code by more than $2,500. That effect is more than 40 percent higher in the area with the Central
Business District (Zone #1) and the Fringe area (Zone #2).
Chapter 3 utilizes the real estate forecasting model to perform a simulation of the effects on
property values in the event of a DHS IBRD anthrax attack scenario used in this study. The chapter first
presents some background on the Seattle economy and on real estate prices. Then, the projected direct
and indirect decline in employment associated with each scenario is entered into model to yield short-run
decreases in property values. These indicate that the scenario used would lead to decline in the median
home price of more than $200,000. Note that the analysis utilizes the basic elements of the attack
scenario of the death of 54,000 inhabitants, which is translated into an employment decrease of 22,000. It
also includes our interpretation of the projected 75 percent decline in the total population of Zone
1stemming from the attack. The negative effect of Scenario 2 on median sales price of homes is $281
thousand, $98 thousand, and $80 thousand, respectively for Zones 1, 2, and 3. Overall the results
indicate that residential property values could decrease by over $50 billion for Seattle, or a 33 percent
overall drop. Moreover, this increases the amount of negative equity by more than $15 billion. This
could result in more than 70,000 residential units being foreclosed.
Chapter 4 presents a computerized spreadsheet model that can be used to analyze the effect of a
disaster on the rate of return on investment for a range of business types. It is based on standard financial
5
methods, but refined for the case at hand. It provides a flexible set of templates for businesses, where
basic characteristics and disaster conditions can be inserted to determine the effect on the bottom line.
The basic characteristics include the economic sector (type of product or service produced), establishment
size, and baseline rate of return. Disaster conditions include the duration of the event and recovery, clean-
up costs and the decline in direct business activity. The chapter presents the basic analytical method and
illustrates it for a variety of business types and sizes, as well as disaster conditions. In general, the
simulation results for Scenario 2 indicate that some Seattle businesses will experience negative rates of
return on assets and equity for several years following the attack.
Chapter 5 provides an overall summary and some suggestions for future research. It again
emphasizes that the methods and results presented here can be used until expert elicitations are
performed. Again, given the fact that even experts have little experience with such extreme disaster
effects, they might serve as a useful cross-check against these and other approaches to determining key
consequences of an anthrax or other major disaster. The results of our analysis are key ingredients into
evaluating the extent of subsidies that would be needed in the event of an anthrax attack on Seattle. Even
our conservative assumption limiting impacts to 54,000 killed (and no significant flight of other
population), would result in mortgage deficits of billions of dollars. Various forms of subsidies to
counteract this would represent a significant expenditure of state or federal funds. Added to this is the
prospect of additional tens of billions of dollars of subsidies to compensate businesses in downtown
Seattle so that they remain viable during the recovery process.
REFERENCES
Giesecke, J., W. Burns, A. Barrett, E. Bayrak, A. Rose, P. Slovic and M. Suher. 2011. "Assessment of the
Regional Economic Impacts of Catastrophic Events: A CGE Analysis of Resource Loss and Behavioral
Effects of a Radiological Dispersion Device Attack Scenario," Risk Analysis, forthcoming.
John, R., H. Rosoff, and W. Burns. 2011. _________________________________________________.
Rose, A. 2009. "A Framework for Analyzing and Estimating the Total Economic Impacts of Terrorist
Attack and Natural Disasters," Journal of Homeland Security and Emergency Management 6(1): 1-29.
Rose, A. G. Oladosu, B. Lee, and G. Beeler Asay. 2009. "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(2): Article 4.
Rose, A. Economic Resilience to Disasters. 2009. Community and Regional Resilience Institute
(CARRI) Report #8.
1
CHAPTER 2. AN ANTHRAX OUTBREAK IN DOWNTOWN SEATTLE: A SPATIAL AND
LONGITUDINAL ASSESSMENT OF TERRORISM’S IMPACT ON REAL ESTATE PRICES
by
Noah Dormady and Thomas Szelazeki
I. INTRODUCTION
The safety and economic threats posed by an anthrax attack to Seattle’s Central Business District
(CBD) and its adjacent zip codes are major concerns to both local and national policymakers. Aside from
the casualties that the attack will produce and the economic downtown that it will engender, the effects on
employers and real estate will also be devastating. Based on the uncertainty behind such a contamination
we expect that hundreds of buildings and large-scale areas will undergo quarantine and clean-up for an
extensive period of time. These effects will undoubtedly pose the decision to a business of whether to
relocate and to a homeowner of whether to abandon, foreclose, or sell a home. It is through this lens that
we analyze the aftermath of such an attack to determine how employment displacement and an economic
downturn will impact housing prices.
The study of the effects of terrorism on property values has been largely absent from literature.
The analysis presented here will develop a model that has the ability both to forecast residential property
values under general economic and housing market conditions and to estimate changes in these values
stemming from employment, population and housing market shocks from terrorism. The impacts of a
terrorist attack on real estate prices will influence the decisions of businesses, homeowners, local
policymakers and urban planners alike. These parties will be able to use the analysis as an instrument to
develop resiliency and expedite the recovery process. Incentivizing businesses and households to return
to the CBD will be a priority. A CBD firm’s ability to maintain functionality, for example,
decontaminating an office building quickly and efficiently, is integral to an expedited recovery process,
which will reduce adverse effects, workforce migration, and decreases in real estate prices. The
magnitude of this disaster will in part be defined by how quickly employment and home prices return to
their pre-disaster levels.
2
II. LITERATURE REVIEW
The effects of terrorism on housing markets have not been widely analyzed, aside from Redfearn
(2005). By examining studies related to natural disasters, we identify parameters and approaches that can
be adapted to a model that captures the economic and housing effects of a terrorism event.
There are various studies that examine how housing markets have responded to natural disasters,
in particular, Hurricane Andrew. Relying on repeat home sales transactions, Smith and Hallstrom (2005)
developed a model that compared how Hurricane Andrew influenced housing prices in Lee County,
Florida, situated adjacent to the Hurricane’s direct path, with counties in the path. They developed two
regression models that used repeat sales prices as the dependent variable. The first model tested home
prices against coastal amenities, which Smith and Hallstrom defined in terms of distance from the coast
and increased risk to damage due to a home’s proximity to that coast. The second model was a hedonic
equation that used the characteristics of the property and the existence of information about a storm’s
potential impact to a home. It was intended to determine whether these two parameters influence
behavior and are sensitive to the treatment of property attributes. The authors found that housing prices
declined by 19 percent in specific flood hazard areas, which implied that homebuyers and sellers are
influenced by information or events in the areas adjacent to directly impacted areas, thereby perpetuating
a stigma.
Smith et al. (2006) further developed the aforementioned work in a study that used simple
regression models to analyze how people and housing markets responded to Hurricane Andrew in Dade
County, Florida, which received the brunt of the hurricane’s impacts. The study finds that the economic
circumstances and capacity of a household appear to be the most significant factors when analyzing
response. Smith et al. examined what patterns occurred in highly damaged areas and in areas with less
damage. Apropos to Smith and Hallstrom, it was found that since the rate of home value appreciation is
slower in flood hazard areas, the population grew faster in these areas when compared to less damaged
areas. In other words, in areas where more than 50 percent of the homes were deemed uninhabitable, the
population grew faster than in areas with less damage. This behavior was driven by financial capacity,
where lower-income households moved into highly-damaged areas due to affordability, and middle-
income households moved out of those same areas because of their financial ability to do so. Since home
abandonment scenarios are present in the Seattle anthrax attack, the impact that stigma has on areas
adjacent to and within target areas is relevant in this case when developing a distance parameter.
Another study that focused on Hurricane Andrew was by Zhang and Peacock (2009), who
identified how quickly home values recovered to their pre-disaster levels. The study determined that
housing sales increased in the wake of the storm, though the rate of appreciation for home values slowed
significantly in the years following. The study focused on the rate of abandonment by homeowners and
3
how this, along with decreased home values, influenced the high volume of sales. This study is of
particular interest not because of its results, which were revealing in themselves, but because of what the
authors mention in their limitations section. The authors state that studying a longer time period before
and after the storm would have produced a stronger model.ii
Superfund sites and their impacts to property values have been widely examined. Aydin and
Smith (2008) studied the effects of post-remediation on property values in general, but more specifically
in Houston. It attempts to determine whether home prices recover after private firms or government
agencies have completed efforts to remove all or most of the hazardous materials. The study’s results are
compared with Kohlhase’s (1991) results as well. Some of the key distinctions between the two studies
were the geographic radius of property value diminution, direct depreciation of property values within
that radius, and the use of more widely available data.iii In Aydin and Smith, the radius where properties
depreciated fell within 3.7 miles of the remediated areas, whereas the Kohlhase study estimated the radius
to be 5.3 miles. The direct depreciation associated with the proximity to the focused remediated site
between 1985 and 1990 is also less, at 8.7% vs. 14.9%. As of 2000, the direct impact on home prices
declined from an 8.7% decrease in 1990 to a 2.2% decrease from the original value of the property (Aydin
and Smith, 2008). The direct impact here is determined by whether or not a property falls within the
estimated radius of diminution. Any effects outside of this buffer are deemed indirect, and interestingly
had the greatest impact on property values and recovery in the area as a whole. It is our presumption that
this is related to a stigma effect. By 1990, for instance, the depreciation in values that was attributable to
indirect demographic effects was 84% (Aydin and Smith, 2008). This diminution was largely attributed
to changes in neighborhood characteristics, where it is common for lower-income households to move
into an area that has been remediated, due to the lower property values. This in turn depreciates property
values.
By 2000, the overall diminution from the indirect demographic effects rose to 93%, and the
author’s deduced that post-remediation, the direct effect, impacted property values less and less over time
and shifted the focus away from the influence of proximity effect to the influence of demographics.
Ultimately, the study finds that preexisting hazardous sites have a tendency to attract higher-income
households, who then abandon the areas after a hazardous event due to their ability to do so. This
demographic is then replaced by lower-income households whose marginal values of a clean, riskless
environment are less and can devalue property (Aydin and Smith, 2008). Stigma tended to drive away
higher-income households, and the authors conclude that, after 20 years, the remediated site has yet to
achieve its full property value recovery.
All of the aforementioned studies used some type of distance variable excluding Zhang and
Peacock, and Smith. Most studies use distance as a continuous variable that is merely the distance from
4
some focal area. Smith and Hallstrom define it as coastal amenity. Kohlhase, and Aydin and Smith
identify it at the point from where property values first change and Redfearn disaggregates distance into
three areas. Redfearn produced a series of models where distance variables were either continuous, or
discrete at three different proximities from a perceived terrorist attack. The models attempted to identify
the idiosyncrasies within each zone and whether property values and perception of risk change with
distance from a terrorism event. Using various highly active locations in Los Angeles as terrorist targets,
he found that risk perception had no adverse affect on the housing markets.
In most studies, the distance variable acts as a proxy or indicator of how people assess risk, which
eventually can influence property values. All of these studies essentially conclude that some combination
of demographic, economic, and distance indicators to a specific area influence property values. In our
model, we seek to explain why and how these variables are induced by a terrorism event at the micro-
level, whose epicenter is a Central Business District.
III. DATA UTILIZED FOR PANEL ANALYSIS
A. Unit of Analysis
Terrorist attacks affect real estate values both through macroeconomic and housing market
variables. We begin with a discussion of a unique dataset consisting of spatially-differentiated
macroeconomic variables.
The sales price of any single property can be driven, to a large extent, by exogenous factors
specific to that property; use requirements of the buyer, preferences, tastes and characteristics inherent to
the location (e.g., square footage) or community in which the property is located. These exogenous, unit-
specific influences on real estate prices, although significant, can detract from the influences of larger
community and regional drivers of real estate prices. It is these larger community and regional drivers
that would be altered by a region-specific terrorism event. For example, a terrorism event such as an
anthrax attack would not alter the square footage of a home, or its number of bedrooms. As a result, we
utilize zip code as our unit of analysis because it enables us to make both community-specific and region-
specific assessments of the macroeconomic drivers of real estate prices, while avoiding the large quantity
of property-specific and exogenous information that may otherwise be redundant in an estimation model.
Zip Codes are a nearly-constant geospatial division of land, used by the U.S. Postal Service for
purposes of mail and parcel management, and provide for both a constant and finely-grained analysis,
while at the same time avoid many of the biasing features inherent in other geospatial locators, such as
Congressional districts or Census tracts. Although there are 61 zip codes within the City of Seattle, our
analysis focuses on only 32 of those zip codes. The remaining zip codes were disqualified from this
5
analysis as they represent “unique” zip codes used by a specific company or organization with a high-
volume of mail service (e.g., a large university), or post office boxes carrying their own zip codes. Eight
zip codes were also omitted because of any one of the following: they fell within a larger zip code, they
were occupied by an airport, university, or federal land, or neither residential characteristics nor data were
present within that zip code.
Our unit of analysis for time scale is months, from January of 1994, through December of 2010.
In addition to purposes of data reliability and ease of access, our historical range of time is theoretically
important because of the region-specific history of the Seattle Metro Area. In 2001, the Nisqually
Earthquake hit the Puget Sound region, shocking the regional economy in the short run and having a
minor downward influence on real estate prices. Our historical range of time, therefore, incorporates time
periods before, during and after that disaster, and also encompasses events such as recessions.
B. Data Acquisition
The data used for this analysis come from a variety of both public and private sources. Our
dependent variable (Median Sales Price) and two other key variables (Quantity of Homes Sold and
Quantity of Foreclosures) were acquired from the real estate and database firm, DataQuick. They acquire
their raw data directly from records available through the King County Department of Assessments, in
Washington State. Each of those variables is provided at the zip code level by month.
The remaining data were available only on an annualized basis. Housing Inventory and
Population were acquired through Seattle’s Office of Financial Management and The Puget Sound
Regional Council (PSRC), respectively. PSRC data were acquired only at the census tract level. We
adjusted this data into zip code levels using the University of Missouri Census Data Center’s Geographic
Correspondence Engine. Furthermore, we were able to generate Population Density using the U.S.
Census’s Geographic Information System (GIS) tool OnTheMap.iv
Employment, and Median Household Income were gathered from The Sourcebook of Zip Code
Demographics, which has been published annually (excluding the year 2002) by CACI International Inc.,
and as of 2003, by Environmental Systems Research Institute (ESRI). We capture employment by place
of work to assess conditions inherent to a locations’ economic climate. Median Household Income
provides for a robust assessment of macroeconomic conditions within each respective zip code.
6
TABLE 1A. DESCRIPTIVE STATISTICS- CITYWIDE Variable Mean Std. Deviation Min Value Max Value
Median Sales Price 322,594 115,631 78,587 1,594,363
Median Household Income 58,253 13,957 17,228 94,162
Employment 13,450 12,693 880 74084
Personal Income ($B U.S.) 148.26 18.07 114.89 172.70
Homes Sold 38 21 1 201
Foreclosures 1.02 1.91 0 22
Housing Inventory 11,513 4,196 4428 23,957
TABLE 1B. DESCRIPTIVE STATISTICS- CONTAMINATION ZONE (ZONE 1) Variable Mean Std. Deviation Min Value Max Value
Median Sales Price 378,681 142,624 91,873 1,594,363
Median Household Income 51,703 18,376 17,228 94,162
Employment 22,954 18,054 3,422 74084
Personal Income ($B U.S.) 148.26 18.07 114.89 172.70
Homes Sold 28 17 1 110
Foreclosures .58 1.17 0 10
Housing Inventory 10,512 2,750 5,416 17,029
TABLE 1C. DESCRIPTIVE STATISTICS- FRINGE CONTAMINATION ZONE (ZONE 2) Variable Mean Std. Deviation Min Value Max Value
Median Sales Price 328,018 107,262 119,675 753,314
Median Household Income 59,189 10,614 42,237 85,284
Employment 12,294 9,446 3,624 55,614
Personal Income ($B U.S.) 148.26 18.06 114.89 172.70
Homes Sold 41 20 6 201
Foreclosures .97 1.73 0 17
Housing Inventory 12,202 4,421 7,429 23,957
TABLE 1D. DESCRIPTIVE STATISTICS- OUTLYING ZONE (ZONE 3) Variable Mean Std. Deviation Min Value Max Value
Median Sales Price 289,426 89,3601 78,587 630,869
Median Household Income 61,186 11,663 25,655 90,208
Employment 9,074 7,943 880 34,770
Personal Income ($B U.S.) 148.26 18.06 114.89 172.70
Homes Sold 40 21 2 164
Foreclosures 1.27 2.26 0 22
Housing Inventory 11,635 4,580 4,428 22,099
7
Of the available seventeen publications of The Sourcebook of Zip Code Demographics, four years
of data were either unavailable or omitted. To account for this missing data, we used a smoothing
technique known as Double Exponential Smoothing to generate an inter-period forecast from the
available data, and utilized those “smoothed” observations to replace missing observations in the original
data. Our smoothed values had an overall fitness (R-squared) of 0.94 and 0.95, for employment and
median household income, respectively.
Finally, we provide Personal Income as a control variable that is measured at the regional level
for the entirety of the City of Seattle. This data are acquired from the U.S. Bureau of Economic Analysis.
This data provides us with a highly robust assessment of the general overall macroeconomic climate of
the region. Median Household Income and Personal Income are measured in constant 2011 dollars,
based on the Consumer Price Index (CPI) for the Housing Sector.
C. GIS Spatial Disaggregation
Our analysis disaggregates the City of Seattle into three specific zones, with zip code 98101 as
the epicenter, whose designation was influenced by the anthrax scenario presented by 2010 Interagency
Biological Restoration Demonstration (IBRD) to the U.S. Department of Homeland Security. See
Chapter 3 for greater details of this scenario and its applications for analysis.
It suggests that an anthrax attack on the Seattle business district would contaminate an area of
approximately 10 square miles. We model these areas spatially using the U.S. Census’ OnTheMap GIS
tool, to create two buffers and generate three contamination zones throughout Seattle. These zones are: 1)
Contamination Zone (includes the Central Business District), 2) Adjacent/Fringe Zone, and 3) Outlying
Zone (shown in Figure 1). The Contamination Zone includes the IBRD assessment area and is
approximately of a two-mile radius, and the Fringe Zone includes a four-mile perimeter surrounding the
epicenter.
8
Figure 1
IV. APPLIED METHODOLOGY OF THE PANEL ANALYSIS
The dependent variable of this analysis is Median Sales Price of residential real estate. It is
measured at the zip code level, and varies monthly across our 17 year panel. The data for this variable are
presented in Figure 1 below. The complexities of analyzing multiple geospatial regions across months
necessitates a model slightly more complex than a pooled cross sectional model, such as a pooled
Ordinary Least Squares (OLS) model. These models do not control for time dependence which leads to
underestimation in the error termv.
A more appropriate method for analyzing these data is the use of a time series cross sectional
(TSCS), or longitudinal model. These models enable unbiased estimates of multiple units across time,
and do not fall prey to the same criticisms as pooled models applied to panel data. The analysis provided
here uses a Fixed Effects Cross Sectional Time Series (FE) model. The model includes a time-stationary
error term that corrects for time dependency, and enables the same style of estimates that are sought
through pooled models or ordinary OLS cross sectional non-panel models.
9
FIGURE 2. SEATTLE RESIDENTIAL MEDIAN SALES PRICE (BY ZIP CODE)
The data generating process for the fixed effects model is given by:
zt zt z zty x u
where:
zty is Median Sales Price.
where 1 z are:
Median Household Income,
Employment (by place of work),
Personal Income,
Quantity of Homes Sold,
Quantity of Foreclosures,
Housing Inventory,
0
500
00
01
00
00
00
150
00
00
Med
ian
Sa
les P
rice
(2
01
1 D
olla
rs)
January 1995 January 2000 January 2005 January 2010
Month (t)
(by Zip Code)
Seattle Residential Median Sales Price
10
where:
zu are unit (zip code) specific fixed effects/ unobservables,
and where:
zt is the identically and independently distributed (i.i.d.) error term.
The regressors included in this equation estimate are both the fundamentals of the local and
regional macroeconomy as well as the fundamentals of the local and regional housing market. Because
the analysis is driven by the aim of estimating spatially-disaggregated effects, this estimating equation is
broken down into five separate regression models. First, the regression is applied to the entire City of
Seattle. Secondly, the model is applied to each of the three “zones” of analysis, which include the
Contaminated Zone, the Fringe Zone surrounding that district, and the outlying area of the city. Finally, a
separate regression is provided which includes only those zip codes that lie outside of the Contaminated
Zone, for purposes of comparison.
V. PANEL ANALYSIS RESULTS
The panel analysis results are provided in Table 1. The regressions are overall robust and explain
anywhere between 35 and 75 percent of the variance in median residential sales prices between 1994 and
2010 for the City of Seattle. The models have strong fitness measures, indicating that a proper set of both
macro and real estate variables have been included. The regression model labeled “Citywide” indicates a
regression that includes each of the 32 zip codes for the entire Seattle area. The models labeled “Zone #1,
Zone #2 and Zone #3” indicate individual zone-specific regression models. And, the model labeled
“Zone 2 & 3” indicates a combined model that includes the fringe zone and the surrounding communities,
excluding the Central Business District/contaminated zone.
The Citywide regression explains the average tendency of real estate prices throughout the City of
Seattle, irrespective of geospatial location. Throughout the panel period, Median Household Income
provides perhaps one of the strongest macro drivers of real estate prices, as it does throughout individual
zones as well. Holding all other variables constant at their mean, a $1,000 increase in median household
income would increase the median sales price of residential property in the mean zip code by more than
$2,500. That effect is more than 40 percent higher in the area with the Central Business District (Zone
#1) and the Fringe area (Zone #2). In those districts, the median sales price of real estate is much more
11
TABLE 1. REGRESSION ANALYSIS RESULTS
Citywide Zone #1 Zone #2 Zone #3 Zone 2 & 3
Median Household
Income
2.58***
(16.76)
3.71***
(7.39)
3.91***
(13.61)
1.65***
(13.11)
2.14***
(18.42)
Employment 0.68**
(3.11)
1.77***
(3.56)
-0.26
(-1.17)
1.49**
(2.88)
-0.27
(-1.36)
Personal Income
($USD Billion)
2982.83***
(44.89)
3108.16***
(12.57)
3119.14***
(29.98)
2755.75***
(47.68)
2860.35***
(57.08)
Home Sales 380.71***
(6.90)
54.29
(0.30)
373.23***
(5.79)
504.11***
(9.49)
475.03***
(11.58)
Foreclosures -5421.85***
(11.36)
-4418.90*
(-1.90)
-5172.96***
(-7.33)
-5394.87***
(14.62)
-5663.79***
(-17.02)
Housing Inventory -6.19***
(-3.71)
-15.83***
(-4.09)
-14.71***
(-4.83)
3.72
(1.53)
3.03*
(1.74)
Constant -217052.39***
(-16.45)
-151144.50***
(-6.05)
-193556.11***
(-8.07)
-295781.27***
(-12.44)
-29980.67***
(-19.11)
N N = 6528 N = 1836 N = 1836 N = 2856 N = 4692
R2 R2= 0.52 R2= 0.35 R2= 0.75 R2= 0.68 R2= 0.71
F F = 1173.05*** F = 162.65*** F = 917.65*** F = 990.06*** F = 1864.68***
Cross-sectional Time Series Fixed Effects panel model results generated using Stata 10. t-values in parentheses.
* p > .1, ** p>.05, *** p>.01.
responsive to changes in household income. Household income is also a highly statistically significant
driver of real estate prices, safely rejecting the null hypothesis that the asymptotic relationship is zero at
even the most stringent levels of significance (p-value) across all geospatial parameters.
The employment rate however, is a strong, but less statistically significant driver of real estate
prices throughout this panel. In those regression models in which the coefficient is statistically
significant, the employment rate is positively related to real estate prices. In the average Seattle zip code,
for every 1,000 job increase in a zip code’s total employment, the median sales price of residential real
estate in that zip code increases by over $680. Recall that Employment is not a measure of the percentage
rate of employment in our panel, but rather provides the total sum of full-time equivalent employed
persons by place of work.
Personal Income, which is an aggregate measure of the strength of the regional macroeconomy, is
shown to be a highly robust and significant determinant of real estate prices. For every $1 billion dollar
12
increase in regional personal income, the average Seattle zip code’s median sales price increases by
nearly $3,000. That relationship is slightly stronger in the Contaminated Zone #1) and fringe
communities (Zone #2), as the same effect leads to a more than a $3,100 dollar increase in median sales
price. Put another way, across our 17 year panel, the mean regional personal income is just below $150
Billion. A single standard deviation increase in the personal income ($18 Billion) would lead to the
average zip code’s median sales price increasing by more than $52,000, holding all other variables
constant at their mean.
The real estate measures also provide for robust drivers of changes in real estate prices. The
variable for Home Sales, which is a measure of the strength of the real estate market, particularly the
demand side, is robust, significant, and positively related to sales price in all of our regression models,
except Zone 1. The Central Business District constitutes the downtown area, and housing prices in those
areas are most likely not affected by changes in demand patterns for housing. Demand for housing in
those communities is probably fairly inelastic, and real estate agents are more than likely, able to find
buyers for residential vacancies who are willing to pay downtown prices. Depending upon the geospatial
location, for every additional property sale in a zip code in a month, the median sales price can increase
between $50 and $500.
Foreclosures are also a robust indicator of the health of local and regional real estate markets.
Foreclosures signal to other home owners that the demand for housing may be in decline, and are often
both a cause and a consequence of a depressed market. Homes that are foreclosed may also be sold for
significant decreases in sales price, particularly if a “short sale” is made, or if a repossession or bank
ownership occurs. As with the other housing market regressors, the median sales price in the Central
Business District is less responsive to foreclosures. Similarly, the mean monthly foreclosure rate of an
average Zone 1 zip code is just above 1 foreclosure per month, whereas that number outside of that zone
is just over 2 foreclosures per month.
The supply side, however, provides a different effect altogether. Housing Inventory provides our
proxy for the supply of available residential real estate. Ceteris Paribus, an increase in the supply of
available real estate should lead to a decrease in the price of real estate, relative to a fixed demand for
housing. We find this effect to be accurate and in the appropriate direction in each of the regressions for
which the coefficient in statistically significant. For every additional residential property added to the
housing inventory in the average Zone 1 zip code, across our panel, the median sales price of housing
decreases by approximately $15. Interpreted another way, a single standard deviation increase
(approximately 2,800) in residential homes would decrease the average Zone 1 zip code’s mean sales
price by over $44,000.
13
The supply of housing provides an interesting departure from expected disaster outcomes. That
is, under a non-disaster scenario, an increase in the housing stock relative to a fixed housing demand
would lead to a decline in the price of housing. However, under a disaster scenario, a decline in the
housing stock stemming from property damage or condemned properties would constitute a decline in
both the supply and demand for housing. There would be fewer livable homes and fewer people would
prefer to reside in them. Under an anthrax attack disaster scenario, we would expect a likely sign reversal
of this coefficient for housing stock, as the demand for housing in both the attack zone and neighboring
communities would decline both in the short run and long run, and the supply of housing would be
suppressed in the short run, with likely sticky rebound in redevelopment. It should be noted, however, for
all coefficients in each of the regression estimating equations, that the assessments were made based on a
panel of data from periods that did not include terrorism events.
VI. DISCUSSION AND CONCLUDING THOUGHTS
This analysis has assessed the relative influence of both local and regional housing and
macroeconomic drivers of real estate prices, by generating a set of cross sectional time series estimating
equations. The variables included in this analysis suggest many significant and robust drivers of changes
in real estate prices across time, and suggest that the real estate market is influenced by patterns of income
and employment, reflecting the health of the local and regional economy. Similarly, the analysis has
suggested that, controlling for economic conditions, the real estate market is similarly driven by
conditions of supply and demand within the local and regional real estate markets.
As discussed previously, there are a number of factors that may influence the price of houses
individually. These include unit-specific amenities like square footage, crime rate, school district quality,
etc. On an asymptotic level, that is in large samples, the median tendency of real estate prices is
unaffected by these micro level unit-specific effects, and is driven by those factors that we include within
this analysis. We acknowledge that there may in fact be alternative local and regional drivers of real
estate sales prices; however, we suggest that including them may be problematic from both a data
acquisitions standpoint and from an econometric standpoint. Overall, however, the summary and fitness
measures of our estimating equations provide a concise and well-explained picture of real estate prices
within the City of Seattle.
The next chapter will utilize the model provided here to estimate dollar figure changes in
residential median sales prices in the event of an anthrax terrorism attack within the Seattle Central
Business District. A set of macroeconomic and housing related market effects will be applied to the
regression models. Both direct effects to the contaminated area and indirect effects to the surrounding
fringe and outer communities will be assessed at both the immediate and short-run horizons
14
ENDNOTES
i The authors are, respectively, Ph.D. Candidate, School of Policy, Planning and Development (SPPD), University
of Southern California (USC), and Master of Public Policy and Master of Urban Planning Graduate, SPPD USC. ii The study only examined a few months prior to the 1992 storm through 1996 when home values had not yet
returned to their pre-disaster levels (Zhang and Peacock, 2009). iii
Kohlhase collected data for the periods 1975, 1980, and 1985. Aydin and Smith collect census data that spans a
wider period from 1970 to 2000 to examine demographic changes. iv Available at: http://lehdmap.did.census.gov/
v The interested reader should see Cameron and Trivedi (2010) for more details.
REFERENCES
Aydin, R. and Smith, B. (2008). "Evidence of the Dual Nature of Property Value Recovery Following
Environmental Remediation," Real Estate Economics 36, 4: 777-812.
Bertelli, A.M. and Carson, J. (2011). "Small Changes, Big Results: Legislative Voting Behavior in the
Presence of New Voters," Electoral Studies, Forthcoming.
CACI Marketing Systems and ESRI. (1994-2010). The Sourcebook of Zip Code Demographics. ESRI
Press: Vienna, Virginia.
Cameron, A.C. and Trivedi, P.K. (2010). Microeconometrics Using Stata: Revised Edition. Stata Press:
College Station, Texas.
DataQuick. (2011). Retrieved from http://www.dataquick.com/
Hallstrom, D.G. and Smith, V. K. (2005). “Market Responses to Hurricanes,” Journal of Environmental
Economics and Management 50: 541-561.
Kohlhase, J. (1991). “The Impact of Toxic Waste Sites on Housing Values,” Journal of Urban Economics
30, 1:1-26.
Office of Financial Management, Washington State. (2011). Retrieved from
http://www.ofm.wa.gov/localdata/king.asp
Puget Sound Regional Council. (2011). Retrieved from http://psrc.org/data/pophousing
Redfearn, C. (2005). “Land Markets & Terrorism: Uncovering Perception of Risk by Examining Land
Price Changes Following 9/11,” The Economic Impacts of Terrorist Attacks.
Smith, V. K., Carbone, J., Pope, J.C, Hallstrom, D., and Darden, M. (2006). “Adjusting to Natural
Disasters,” Journal of Risk and Uncertainty 33: 37-54.
U.S. Census Bureau OnTheMap. (2011). Retrieved from http://lehdmap.did.census.gov/
Zhang, Y. and Peacock, W.G. (2009). “Planning for Housing Recovery? Lessons Learned from Hurricane
Andrew,” Journal of the American Planning Association 76, 1: 5-24.
1
CHAPTER 3. THE POTENTIAL IMPACT OF AN ANTHRAX ATTACK
ON REAL ESTATE PRICES AND FORECLOSURES IN SEATTLE
by
Noah Dormady, Thomas Szelazek, and Adam Rose
I. INTRODUCTION
This chapter applies the real estate forecasting model presented in Chapter 2 to the DHS anthrax
attack scenario (AAS). To summarize, the Interagency Biological Restoration Demonstration (IBRD)
Seattle scenario presented to the U.S. Department of Homeland Security, estimates that 108,000 people
would be exposed to acute anthrax poisoning, resulting in 54,000 casualties. These effects would be
localized primarily in Seattle’s Central Business District.
The attack affects property values through several channels. This includes contamination and
temporary or permanent closing of buildings in the area in which the anthrax is dispersed, fear of
contamination in a wider area, concern about longer-term contamination/distrust of the effectiveness of
clean-up efforts, general economic decline, and factors specific to the housing market. Note that our
analysis is confined to residential property values. The effect of an anthrax attack on business property
values can be projected with an analogous model of the commercial and industrial real estate markets.
The basic simulation results can be examined in isolation to evaluate potential offsetting government
assistance or can be entered into the Financial Model presented in Chapter 1 to evaluate assistance in the
context of rate of return subsidies.
At the end of the chapter we combine the real estate price projections with data on Seattle
housing stock to estimate total declines in real estate prices and the number of people whose mortgages
would probabilistically be underwater. This provides useful information on potential defaults. The
results can be used to evaluate government funding needs to avoid dire negative outcomes of the anthrax
attack.
II. BACKGROUND
The center of the anthrax attack occurs in Seattle’s Central Business District (CBD),
contaminating approximately ten square miles. We designate this as the Contaminated Zone (Zone 1).
2
We term the neighboring area as the Fringe Zone (Zone 2), and the remaining area of the City as the
Outlying Zone (Zone 3). Due to the differing population and employment densities of the zones, we
expect that the housing market zones will suffer independent impacts from one another. For example, the
Contaminated Zone has approximately 22 percent more employees than residents. The area is largely
made up of businesses, which collectively employ over 220,000 people. The impact of the anthrax attack
on this area’s real estate market could potentially depreciate property values more than the residential
Outlying Zone, because it is heavily dependent on companies’ ability and decision to return.
The condition of the housing market is largely determined by economic and demographic factors,
both of which influence housing demand. Employment growth, unemployment rate, and formation of
new households are all key indicators when evaluating the health of a housing market. Many of the
226,000 people, who work in the Contaminated Zone (Zone 1), will be affected in a manner that will alter
their employment. Whether it is through relocation, severance, or work from home situations, the
housing market in the region will be impacted by Zone 1 residents and the specific labor force that
commutes to work from other zones. The impact on real estate from the attack in the CBD area will be
determined in part by the magnitude of business and employment disruption. Normally in cases where
natural or man-made disasters impact a region, the area at the center of the disaster suffers the brunt of the
direct effects (i.e. demographic displacement, home abandonment) while, the fringes experience indirect
effects (as a result of both accurate information and exaggerated information associated with stigma that
leads to property value depreciation). Interestingly, areas that experience direct impacts from disasters
typically recover quicker than the outlying areas. In events such as hurricanes for instance, the homes
that lie in the path of the storm generally end up being abandoned and appreciate at a lower rate than in
normal weather conditions. This slower rate is largely attributed to information and stigma effects. The
demand is filled after a period of a couple of years, but the area’s property values suffer because
households with lower median incomes are attracted by the lower home sale prices (see, e.g., Smith,
2006).
In an area where land remediation is required, such as the experience with superfund sites, a
buffer area arises out of uncertainty and fear of the spread and/or lack of thoroughness of the clean-up
effort. , This would cause a spillover of the impact on real estate values. Depending on the radius of the
buffer, the direct effects of decreasing property values will be influenced by demographic changes in the
area. However, the real change could come from indirect effects such as information, stigma, job
relocation, and geography. Depending on the timeliness of clean-up efforts and the recovery process, all
three zones will suffer from these indirect effects thereby increasing vacancies and slowing any rebound
in real estate prices. . Apartment and office space vacancy could potentially suffer the most, as the target
area is very dense in employment. In Houston, for example, layoffs during the events of the Enron
3
Scandal and 9/11 impacted downtown real estate significantly leaving many vacancies. The potential
impacts to Downtown Seattle real estate could have even more lasting effects due to its concentrated
population and employment cluster. Even if the clean-up is successful, residential and commercial
property values could decrease more than the 12 percent decline that resulted from the recent recession
between 2009 and 2010 (Conway, 2010). In such an event, demand is normally not met by returning
residents and employees, but by a new demographic that will be enticed by lower housing prices and
work opportunities.
In the case of the two outer zones, we expect that the indirect effects, such as stigma, would play
a major role. Depending on socioeconomic status and insurance policies, home abandonment could
become prevalent as people leave the area in search of homes elsewhere, which would additionally
exacerbate real estate stigma. Furthermore, housing recovery could potentially be slower than in a
downtown zip code, which is part of a regional economic hub. All three zones have different ways of
attracting a household. In the outlying zone, the demand is met by single family household supply, while
in downtown it is largely the type of businesses and employment that drive housing demand, mostly in
the form of condominiums and apartments. In the wake of the attack, the housing market will begin to
become healthier once businesses return and employment is regenerated. In the meantime, many
employees will relocate, quit, incur severance, or work from home. As time passes and the clean up
continues, the behavior of these employees is what will ultimately decide the health of the housing
market.
The fact that the average commuting distance for a Seattle worker is twelve miles suggests that
many people who work in a zip code normally do not live in that zip code, as made evident by the
employment- population disparity in the CBD. This is important to consider when estimating the impact
on property values, because one area’s loss may mean an exodus to another area in the Puget Sound.
While the CBD in Seattle may suffer depreciation in its housing market, one of its neighboring areas may
experience an appreciation. To estimate this effect, it is essential to distinguish between the direct effects
of the attack and remediation, and the indirect effects of employment shifts, business relocation, and
stigma effects.
Additional insight into the effects of a disaster on property values stems from the scandal and
collapse of Enron in Houston in 2002. This caused a significant increase in office vacancy rates in the
metropolitan area. In fact, the Class A office vacancy rate in downtown rose from 2 percent before the
event to 16 percent in 2003. Still, the effects of the Enron collapse were not as great as the 1980s oil bust,
where the Houston area lost 250,000 jobs (Woodyard and Kasindorf, 2002). Interestingly, there was no
significant increase in home foreclosures due to the bankruptcy. The main reason was that many of the
people who had worked for Enron were able to quickly begin work with another company and thus avoid
4
major personal financial troubles (Ahrens, 2011). One problem with comparison to the Seattle anthrax
attack is that, in the Enron case, employees stayed in the city and were able to effectively find another
job-. Another major difference is that Enron mostly occupied Class A office buildings. In the aftermath
of a terrorist attack, meanwhile, Seattle would suffer increased vacancy rates in all classes of office
buildings as well as in home, retail, and industrial markets.
III. THE SEATTLE ECONOMY
The Seattle metropolitan statistical area (MSA), as defined by the United States Census Bureau, is
comprised of the Counties of King, Snohomish, and Pierce, and includes the Cities of Seattle, Tacoma,
Bellevue and Everett. With a population of nearly 3.5 million, the MSA is ranked 15th in the U.S. It
makes up the majority of the Puget Sound Region. Puget Sound is a large seaway, or ocean inlet, that has
a complex system of estuaries, bays, and harbors. The three Seattle MSA counties make up the majority
of the eastern side of the Sound, which is nearly two times longer than the west, while Kitsap County
occupies the western side. We offer a description of the economy of the entire Puget Sound Region, with
a focus on the Greater Seattle area, due to the enveloping geography of the area and the fact that the
Sound itself acts as one large port to a vast array of industries and businesses bordering it.
The major pillars of the Seattle Metropolitan Region’s economy are the manufacturing,
construction and high-tech sectors. Company headquarters of Amgen, Boeing, Microsoft, Nordstrom,
Russell, Costco, Amazon, Starbucks and Weyerhaeuser are located throughout the region, attracting a
strong, diverse labor force and generating globally competitive regional output. The manufacturing
industry is highly reliant on the aerospace sectors, as they make up nearly half of the employment of the
industry, with Boeing as the largest employer (74,160 employees) (Conway Pedersen Economics, 2009;
University of Washington, 2010).
The overall regional economy is largely dependent on the health of the manufacturing industry
and the aerospace sectors (Conway Pedersen Economics, 2009). In the latter parts of the past four
decades the aerospace sector experienced patterns of rapid growth in employment due to highly
stimulated national economies. In the 1980s, economic expansion helped create more than 20 percent of
the jobs in the Region today. The recessions of the late 1970s and early 2000s in the Region were largely
attributed to Boeing’s decline during those periods. Conversely, the current recession has been largely
unaffected by Boeing or the aerospace sector.
The information technology industry is made up of nearly 3,000 companies, with employment at
around 206,000 (OnTheMap, 2011), led by Microsoft with over 38,405 employees (University of
Washington, 2010). The region is a hub for many start-up and technology savvy businesses because of
5
the high volume of highly skilled and technical workers. Over the past twenty years, high-tech companies
in the fields of internet, computers, design and clean technology have diversified the economy. This
diversification has decreased the employment share of goods-producing industries from 23 percent in
1990 to 16 percent in 2009 (Conway Pedersen Economics, 2009).
In addition, the Ports of Seattle, Tacoma and Everett are major economic engines as they are
modes of ingress and egress for facilitating the distribution of products from all sectors in the region.
Collectively, the Ports are the 4th busiest in North America (American Association of Port Authorities,
2009). Albeit highly active in oversea transactions, the Ports’ biggest importer is the U.S.
Healthcare and biotechnology sectors, ranked 5th in the U.S., are also significant contributors to
the regional economy, employing over 192,000 people and contributing $10 billion annually to the gross
regional product (City of Seattle, 2007). The two sectors’ contribution to the regional economy has
partially stemmed from grants by the Bill and Melinda Gates Foundation, which is locally based.
Contributions have also propelled the University of Washington to becoming the City of Seattle’s largest
employer and the Puget Sound Region’s third largest employer, with nearly 28,000 faculty and staff. In
addition, it produces a direct economic impact of $3.7 billion and total impact of $8.6 billion (University
of Washington, 2010). The University’s medicine branch is also a key contributor as it creates an
additional 16,000 jobs and generates an additional $1.7 billion directly.
Finally, the military has a large presence in the Puget Sound region in terms of Army, Naval, and
Air Force bases. Major facilities, such as Fort Lewis Army Base and the Puget Sound Naval Shipyard in
Bremerton, are collectively one the largest employers in the region.
IV. CURRENT ECONOMIC AND HOUSING INDICATORS
Prior to the inception of the current recession, the Region was experiencing employment and
population growth of more than twice the national rate. However, because the region entered the
recession one quarter later than the nation, it began losing jobs faster than the national average due to the
lag. The employment erosion was exacerbated by the failure of Washington Mutual, and layoffs
conducted by Boeing and Microsoft. Tables 1.1 and 1.2 exhibit the differences at the local and national
level in percent changes.
6
Source: Puget Sound Economic Forecaster, 2010.
Figure 1.1 Puget Sound and U.S. Employment
Annual Percent Change
Source: United States Census Bureau, 2010
Figure 1.2 Puget Sound and U.S. Population
7
From the first quarter of 2008 to September 2010, the region experienced an employment loss of
7.1 percent (131,400 jobs) compared to a 5.8 percent loss for the rest of the nation (Conway and Pedersen,
2011). The construction industry, which employed over 125,000 people in 2008, has been the most
affected by the recession, with employment at approximately 90,000 today. These cuts have largely led to
the drastic decline in housing permits issued.
As of June 2011, employment is 1,706,090, and the unemployment rate is 9.2 percent. Personal
income as of 2009 was $171,680,771 and per capita stood at $50,378 (Bureau of Economic Analysis,
2010). Comparatively, the per capita income average for 365 metropolitan areas designated by the
Bureau of Economic Analysis is $40,757 (the Seattle MSA ranks 15th)
. This high ranking is a result of a
strong economic influence of the region’s information technology and manufacturing industries, which is
accompanied by a skilled and diverse labor pool. Because of this labor pool, Seattle’s MSA output had
fully recovered in September 2010 by growing 3.4 percent more than the third quarter of 2008, its pre-
recession peak. This growth resulted in an 8th place ranking among the top 100 metro areas (Wial and
Shearer, 2009).
The strength of the regional economy prior to the start of the recession propelled activity in the
housing market to unprecedented levels. Home sales and construction grew to levels higher than most
parts of the country; therefore, when the housing and credit markets collapsed, the fall of the region’s
economy, being a quarter behind, was delayed and more profound. Like most cities in the U.S., Seattle
was responding to a demand in the housing market, which eventually morphed into a recession. The
Region’s recessions in the late 70s and early 2000s were engendered by a reduction of demand for
regional exports, mostly stemming from Boeing’s problems (Conway and Pedersen, 2010).
When the speculative housing bubble burst, Seattle like the rest of country experienced record
low rates of home building. Within a year, regional home sales fell 63 percent, residential building
permits were down 51 percent, and housing prices declined by 19 percent, all of which ultimately
contributed to a loss of 43,500 construction jobs (33 percent of total job loss) (Conway and Pedersen,
2011). Two key differences separated the Puget Sound Region from the nation in regard to the effects of
the finished homes that were constructed during the bubble: the Seattle area’s land and building
regulations. Also, the City’s topography prevented overbuilding during the boom, something that was
prevalent in other parts of country, like the Sun Belt region.
Since the first quarter of 2009, average home sales prices have decreased by 8 percent ($365,558
to $336,284); however, the housing market appears to be on an upside trend. Homes sales have increased
by 9 percent since 2009, and the rental market is performing strongly. According to Dupre + Scott
Apartment Advisors, the apartment vacancy rate has dropped from 7.1 percent to 4.6 percent from the
third quarter of 2009 and the first quarter of 2011. Average monthly rent bottomed out at $951 in 2010
8
and has since rebounded to $974, another indicator of the increase and preference for renting (Conway
Pedersen Economics, 2011).
The issuance of permits for both single and multi-family homes has decreased by 29 percent from
pre-recession levels. While single-family permits have rebounded almost entirely to pre-recession levels,
multi-family permits, after suffering a 40-year low in 2009, have remained at a depressed level (Conway
and Pedersen, 2011). Last year permits were down over 50 percent, but, due to the vacancy rates hovering
at levels lower than pre-recession, the demand for apartments could facilitate the increase the permit
issuance of multi-family housing.
Taxable retail sales, which was the region’s largest tax base declined by 20 percent during the
recession. Over the past three decades, real per capita taxable retail sales have increased at an annual rate
of 1.2 percent (Conway and Pedersen, 2010). In the third quarter of 2009, however, taxable retail sales
decreased by 17 percent; however since the middle of 2010 total sales have increased by 6.2 percent.
Returns to the pre-recession trend line are not expected until 2014 (Conway and Pedersen, 2010).
Compared to most metropolitan areas, Seattle’s economy has been relatively resilient due to its
appeal to start-up businesses, and its export-orientated and innovation-driven economy. This appeal
largely stems from a growing entrepreneurial community that has spawned in Seattle during the last
decade. Many of the bigger companies, such as Amazon and Microsoft, create opportunities for research
and innovation that are ultimately pursued by small start-ups (Schweitzer, 2010). This investment by
bigger companies has created a hotbed of complementary businesses that support the local economy and
generate higher levels of personal income than most parts of the U.S.
Table 1. Annual Percent Change in Major Economic Indicators
2008 2009 2010 2011
Puget Sound Region
Employment 0.9 -5.1 -1.7 1.3
Personal Income (Cur. $) 3.1 -0.9 2.5 5.0
Consumer Price Index 4.3 0.6 0.3 2.6
Housing Permits -43.0 -49.8 30.1 8.2
Population 1.4 1.5 0.9 0.8
United States
GDP ($05) 0.4 -2.6 2.7 3.2
Employment -0.6 -4.4 -0.7 1.3
Personal Income (Cur. $) 2.9 -1.7 3.1 5.3
Consumer Price Index 3.8 -0.3 1.6 3.1
Housing Starts -32.9 -38.4 5.5 6.5
Population 0.9 0.8 0.98 0.97
Source: Puget Sound Economic Forecaster, 2011.
9
More recently, strength in these dimensions of the economy is being challenged by overseas
markets, predominantly Asian. China’s major advances in research and innovation have engendered a
sense of urgency in metropolitan areas such as Seattle to accelerate innovation into the marketplace and
invigorate technology sectors. With aerospace, exports and information technology largely stable, Seattle
is outperforming many of its national counterparts and its recovery has been stronger than the majority of
the Nation’s.
V. REAL ESTATE IMPACT MODEL
A. Basic Considerations
We utilize the real estate predictive model from Chapter 2 to estimate the effects of two stylized
anthrax attack scenarios on real estate prices in Seattle. These scenarios may be considered from two
broad ranges of potential impacts, where the first scenario constitutes a probabilistic lower bound, and the
second scenario constitutes a probabilistic upper bound of impacts. Our methodology employs the panel
model coefficients additively to determine the net change in median residential sales price by concentric
attack zone. We insert into the model independent macroeconomic and housing market variable values
reflecting the current period (short-term) impacts of the attack. These variables are applied differentially
across the three Seattle geographic zones:
Zone # Zone Name Population Employment
1
Contaminated Zone
176,417
226,127
2 Fringe Area 252,323 116,286
3 Outlying Area 387,371 118,594
_____________________________________________________________________________________
Recall that the key explanatory variables in the real estate predictive model are:
• Employment by place of work
• Median household income
• Total personal income
• Home sales
• Foreclosure
• Housing inventory
10
We simulate the effects of two scenarios:
1. Scenario 1-- Significant Casualty Rate. The DHS Interagency Biological Restoration
Demonstration (IBRD) Scenario projects 54,000 deaths as a result of the anthrax attack. In this
simulation, we use deaths, injuries and direct contamination data as inputs. That is we simply assume a
population decline of 54,000, with no out-migration. The other feature is the closing of the CBD to
economic activity during the decontamination period. This is a very conservative interpretation of the
AAS and thus yields a lower-bound estimate.
2. Scenario 2 -- Major Out-Migration. This simulation includes a 75 percent population flight
from Zone 1. The reduction was deduced from survey respondents’ reactions to the IBRD scenario and a
written script of events that take place in the aftermath of the AAS (see Burns et al., 2011). It should be
noted that this migration is not permanent for a portion of that 75 percent population, who are expected to
begin returning after 3 to 4 months. An out-migration of this magnitude would essentially cause the
City's economy and real markets to implode. No model could accurately predict the decline in real estate
values for this extensive of an event, and hence so great a departure from the historical statistical
database. Suffice it to say, however, that a 75 percent population decline would lead to property value
declines of at least that magnitude. We simulate a modified version of this scenario, which allows us to
illustrate the use of the model in a meaningful way. This is to assume the exit of 75 percent of just Zone 1
population, which represents slightly over 35 percent of the total population in the 3 zones. We consider
this a mid-range estimate of real estate impacts. A simulation of the flight of 75 percent of the population
from Zone 1 alone would represent an upper bound. It is important to note that the ranges between the
population and employment values, particularly for Zone 1, vary significantly between the two scenarios
and greatly influence the results of this analysis.
B. Methodology
Employment is a major explanatory variable. However, the AAS does not specify the
employment loss associated with the attack. In Scenario 2, we utilize the AAS population reduction from
short-term 54,000 deaths as a basis for estimating employment impacts. We assume a fixed ratio of
employment to population in the Seattle area; this corresponds to a constant labor force participation rate
and unemployment rate. We then use population and employment data from the Impact Analysis for
Planning (IMPLAN) System (MIG, 2011) to convert to direct employment impacts in Zone 1. The
IMPLAN I-O Table for King County is then used to estimate the indirect employment impacts throughout
the three zones.
11
The total (direct plus indirect) employment impacts are then entered into the predictive model
equation. Analogously, we convert the 75% population migration into an employment decline for
Scenario 1. However, this involves one major complication, in that more people work in Zone 1 than live
there due to the densely employed CBD. Thus, we assume an employment decline of 90% of base Zone 1
employment for input into the model. The macroeconomic declines, including employment due to the
AAS, are provided in Table 2 below.
In a similar fashion, we use IMPLAN data to calculate the declines in the associated total
personal income for the citywide region. Although IMPLAN provides for more disaggregated personal
income decomposition in a single year, we utilize personal income reduction at the citywide level so that
it will be consistent with our panel analysis, which uses publicly-available region-wide personal income
throughout the panel period. The total citywide personal income reductions are provided in Table 2.
Without providing unnecessary detail, it should be noted that the largest personal income effects come
from changes within Zone 1 in both scenarios. Finally, Median Household Income is determined
Table 2a. Macroeconomic Variable Estimated Changes (54K Deaths Scenario)
Scenario 1 Citywide Zone 1 Zone 2 Zone 3
Employment Change -56,436 -3,543 -3,852
Median Household Income Change
(constant 2011 dollars) -14,841 -2,028 -2,426
Personal Income Change
(constant 2011 dollars) -5.94B
Table 2b. Macroeconomic Variable Estimated Changes (75% Population Flight Scenario)
Scenario 2 Citywide Zone 1 Zone 2 Zone 3
Employment Change -192,379 -12,341 -13,418
Median Household Income Change
(constant 2011 dollars) -50,592 -7,065 -8,453
Personal Income Change
(constant 2011 dollars) -20.69B
12
secondarily from IMPLAN employment outputs, because IMPLAN does not provide Median Household
Income as an output. We determine zone-specific median household income effects proportionately from
IMPLAN’s respective employment outputs.
We also estimate similar scenario-specific real estate market impacts. As is consistent with both
theory and scenario development, we assume that housing inventory will change consistently on the basis
of the quarantined and condemned residential structures (single-family dwellings), as well as multi-family
dwellings that are converted into individual units. Because the differences in our two scenarios are
irrespective of the number of housing units that would otherwise be condemned, we utilize the same
housing stock reduction for both scenarios. We assume that there will be a 50 percent decrease in the
available housing stock, which translates into approximately 5,599 homes in Zone 1. Given that there are
a total of 9 zip codes within this zone, we estimate that the average Zone 1 zip code will see a reduction in
housing inventory of 622 housing units.
The effect of the AAS on housing sales is estimated for each zone by assuming a single standard
deviation decline from the mean monthly quantity of property sales for the average zip code in each of the
three respective zones. Descriptive statistics for these figures are provided in Chapter 2. These single
standard deviation declines are 13, 17 and 15 fewer homes sold, in Zones 1, 2 and 3 respectivelyvi. We
maintain that this provides a conservative estimate of the potential declines in home sales that may
actually occur in the event of an AAS-comparable event.
VI. RESULTS
As mentioned in the previous section, our methodology employs estimates from the IBRD
scenarios to determine levels of both macroeconomic conditions and real estate market parameters ex
ante. We then secondarily utilize these parameters as inputs into the real estate market predictive model
from Chapter 2, to determine a range of possible real estate market effects.
The values provided in Figure 2 summarize the impacts from each of the variable imputations.
That is, they predict a short-run change in the median sales price of residential real estate in each of the
three AAS concentric zones given each of our macroeconomic and real estate market parameters from our
two AAS scenarios. These include the direct (Zone 1 only) and indirect effects stemming from each
scenario’s impact on employment, personal income and median household income. These similarly
include the estimated real estate market impacts stemming from a 50 percent decline in housing stock in
each of the Zone 1 zip codes, and a single standard deviation decline in home sales for each respective
zone.
13
Figure 2. Median Sales Price Change by Concentric Zone (Scenario 1 and 2)
The results suggest that the average zip code in the Central Business District (Zone 1) would see
an immediate decline in the median sales price of residential real estate by over $74,000, under Scenario 1
conditions, and an immediate decline in the median sales price by over $280,000 under Scenario 2
conditions. The Zone 2 and 3 impacts under both scenarios (provided in Figure 2) are significantly
smaller than those in the midst of the attack area. From the macroeconomic side, those outer zone
impacts are driven by indirect effects that influence patterns of employment and economic conditions
from interrelationships between zones within the regional economy. Region-wide personal income
effects provide the largest share of these macroeconomic impacts. From the real estate side, these effects
are driven by the decline in home sales. Home sales declines would likely stem from the demand side,
through both income effects and psychological (stigma) effects that would be cross-pollinated from the
central attack zone throughout the rest of the city.
It is noteworthy that increases in housing inventory have historically provided a supply-side
effect that has led to declines residential real estate prices. Simply put, as the supply of available housing
units increases, the price of housing relative to demand declines. In a post disaster scenario, we estimate
that there would be a proportionally small effect in the opposite direction. That is, condemned housing in
the Zone 1 would decrease the supply of housing units relative to demand for housing, and, thus, there
would be a slight increase in the price of housing. We estimate this increase to be approximately $9,800
for Zone 1. Given the decomposition of these effects for the supply of housing units, however, the
-$75,417
-$32,637 -$28,288
-$280,722
-$98,172 -$79,975
-$300,000
-$250,000
-$200,000
-$150,000
-$100,000
-$50,000
$0
Zone 1 Zone 2 Zone 3
Median Sales Price Change
Scenario 1
Scenario 2
14
negative effects stemming from declines in home sales and macroeconomic changes overwhelm this
small positive increase.
We apply these macroeconomic and real estate variables additively, as in the form of a linear
model. Because our panel model is a cross-sectional/time series model, with monthly variation in the
data, our estimates provide predictive quantification for changes in median sales price for a short-run
change in these respective markets. We maintain that the Scenario 1 price changes can be considered a
short-run lower bound estimate, and the Scenario 2 price changes can be considered a short-run upper
bound estimate. Residential property value losses for all three zones total $15.3 billion and $50.5 billion
in Scenario 1 and 2, respectively (see Table 3). Table 4 presents two sets of residential property value
estimates prior to the attack. The first set reflects a strong pre-recession housing market (2007), while
the second indicates a currently recovering market (2010). However, the two sets of results are very
close. Thus, the housing value declines from a terrorist attack represent a 10 percent drop for Scenario
1 and a 33 percent drop for Scenario 2.
Table 3. Residential Property Value Losses in Seattle from an Anthrax Attack
Scenario Zone 1 Loss ($B) Zone 2 Loss ($B) Zone 3 Loss ($B)
1
$7.251
$3.419
$4.660
2 $26.992 $10.285 $13.176
Table 4. Total Residential Property Values prior to the Anthrax Attack (2011 $)
Zone # Total Value 2007 ($B) Total Value 2010 ($B)
1
$43.652
$45.352
2 $49.454 $45.569
3
$64.164
$54.408
15
Given these estimates, it should be noted that our model excludes medium- to long-run resiliency
in the real estate market and the greater regional economy, as well as the potential likelihood of at least a
small number of the City’s residents returning post-attack. Considering past natural hazards and terrorism
events, both domestically and in other developed nations, it can also be assumed that there would be at
least a modest degree of capital influx and subsidization stemming from relief efforts and political
support, which would have medium- to long-run impacts that are similarly exogenous to our analysis.
VII. FORECLOSURES
The predictive estimation provided in Section VI, as well as in the panel model, explicitly
controlled for the impact of foreclosures on changes in real estate prices. Our predictive estimation from
the previous section controlled for those values by holding them at their historical mean. The estimation
of the panel model coefficients were ultimately taken from historical periods where markets were not
constrained by hazard effects. As provided in Chapter 2, foreclosures can have significant impacts on a
local community’s housing prices. Given this, we provide a rough estimation of foreclosures by
concentric zones, given the predicted sales price changes provided in the previous section.
Our assessment of foreclosures is based upon the assumption that changes in the sales price of
real estate, which ultimately influences property values, would cause a number of housing units to retain a
market value below current market values, and similarly, would increase the number of properties that are
“under water” (have negative equity). We employ national survey data from the US Department of
Housing and Urban Development (HUD) to determine the distribution of homes that would otherwise
have negative equity, given proportional changes in median sales price, as provided in Section VI.
Every two years, HUD conducts a comprehensive regional American Housing Survey (AHS).
Fortuitously, he 2009 survey was conducted in the Seattle Area and included nearly 1,000 respondents.
We use two key variables from the survey. The first is the level of mortgage/home loans (in dollars). The
second asked respondents how much money they have left to pay on that loan (outstanding balance).
From these key variables, we were able to determine a citywide distribution of Homeowner Equity
(money already paid toward a home loan). Our foreclosure analysis assumes that the large sample survey
is asymptotically equivalent to the true distribution of the Seattle regionvii
. Similarly, we utilize this
distribution of equity values to determine the probability that a housing unit within a concentric zone will
have negative equity following a proportional reduction in median sales price, consistent with our
estimates for each scenario and each zone in Section VI.
16
Figure3. Kernel Density Estimation and Quintile Decomposition of Seattle Homeowner Equity
The likelihood of a housing unit foreclosing depends upon the degree to which it maintains
negative equity. For example, a homeowner who owes $500,000 on a home that has a post disaster
market value of $200,000 would have a high probability of foreclosing, relative to a homeowner who has
a higher proportion of equity in the home. As such, we decompose this negative equity distribution into
five quintile ranges. Figure 3 provides a kernel density figure of these quintiles for the Zone 1 Scenario 1
equity distribution. Those housing units in the largest quintile (furthest to the left) have the largest
negative equity, and thus, would have the largest probability of foreclosing. We apply these quintile
ranges to a fixed scale of foreclosure likelihood, as provided in Table 4.
Table 4. Negative Equity and Foreclosure Rate Scaling Parameters
Negative Equity Quintile
Foreclosure Rate
<10 % 0 %
10 -25 % 10 %
25 -50 % 25 %
50 -75 % 50 %
>75 % 75%
17
Each scenario specific median price change is applied to the quintile range of negative equity for
each zone, as given by equation 1:
q
z
T
UU
U
(1)
where
qU is the number of housing units within a negative equity quintile from the
US HUD's American Housing Survey,
TU is the total number of housing units (completed responses) from the
US HUD's Seattle American Housing Survey ( 91 survey responses),
zU is the total number of housing units by concentric disaster zone,
is a scaling parameter equal to Pr( | ( ))F Q u the conditional probability
that a housing unit will foreclose given its percentage of negative
post-shock equity
Our foreclosure estimates are provided in Figure 4. We estimate that there would be
approximately 8,700 housing units in foreclosures in the Zone 1, following an impact consistent with
Scenario 1, and approximately 42,000 housing units impacted by foreclosures in the same zone following
an impact consistent with Scenario 2.viii
The average zip code wide estimates can be derived from these
numbers, by dividing each zone-wide total by the number of zip codes within that zone: 9 zip codes in
Zone 1, and 9 and 14 in Zones 2 and 3 respectively. This accounts for the greater total number of
foreclosures in Zone 3, because that zone has a proportionately larger number of zip codes, encompasses
a larger geospatial area, and has more housing units. Overall, we project nearly 17,000 foreclosures for
Scenario 1 and over 72,000 for Scenario 2.
Given disasters of the magnitude simulated in our scenarios, a large number of housing units
would lose a significant amount of equity due to the declines in sales price. Our analysis also estimates
the total sum of negative equity from all housing units, whether they foreclose or not. For example, a
homeowner who has an original mortgage of $400,000 and has currently paid $50,000 toward that home
loan, and is hit by a sales price decline of $75,000, will wind up holding $25,000 in negative equity. In
other words, that homeowner owes $350,000 on a home he/she can currently sell for only $325,000.
18
Figure 4. Estimated Foreclosures (by zone)
Furthermore, we also take the added step of accounting for extant negative equity. We utilize the
US HUD AHS results for the City of Seattle and assume that the zone-wide distribution of amortization is
asymptotically equivalent to that of each of our three concentric zones. Given the original mortgage
value of each housing unit and current equity in that unit, we calculate the post-disaster equity for each of
our disaster scenarios for each zone. We then separate those housing units into quintiles of negative
equity, and determine the quintile-specific mean dollar value of negative equity. We then multiply that
total by the quantity of housing units that we estimate would be within that quintile of negative equity
following each disaster scenario (based on the assumption that the distribution of equity in each zone is
equivalent to the equity distribution of respondents from the 2009 AHS). Next we aggregate these
subtotals into a zone-wide total of negative equity following each disaster scenario for each zone. Finally,
we deduct from these totals the estimated totals of pre-disaster negative equity. In each zone, that
amounted to approximately 3 percent of the total quantity of housing units, with each of those housing
units holding a mean pre-disaster negative equity of just over $33,000.
The results are presented in Table 6. Total negative equity increases for all three zones by $1.4
billion in Scenario 1 and $15.2 billion in Scenario 2. Negative equity more than doubles over the $0.6
billion pre-disaster level in Scenario 1, and increases nearly 25-fold in Scenario 2. Negative equity is
only about 10% of total value losses for Scenario 1 but about 30% for Scenario 2.
8,767
3,509 4,615
42,211
14,101 15,931
0
5,000
10,000
15,000
20,000
25,000
30,000
35,000
40,000
45,000
Zone 1 Zone 2 Zone 3
Zon
e-w
ide
Qu
anti
ty o
f Fo
recl
osu
res
Foreclosures by Zone
Scenario 1
Scenario 2
19
Table 5. Total Post-Disaster Negative Equity Value (by Zone)
Scenario Zone 1 ($B) Zone 2 ($B) Zone 3 ($B)
1 -$.738 -$.277 -$.365
2 -$11.790 -$1.680 -$1.732
VIII. CONCLUSION
This chapter has illustrated the usefulness of a statistical analysis of the Seattle real estate market
to analyzing some of the more important impacts of an anthrax attack in that city. The model's ability to
predict a decrease in housing prices following such an attack can serve as the basis for estimating the
vulnerability of the real estate market. The model is used to help predict the extent to which Seattle
residents will find their mortgages underwater, and hence offers an indication of the amount of
government assistance that might be needed to prevent defaults/foreclosures and hence a large population
exit in the city. Our results for our worst case scenario indicate that residential property values could
decrease by over $50 billion, or a 33 percent overall drop. Moreover, this increases the amount of
negative equity by more than $15 billion. This could result in more than 70,000 residential units being
foreclosed. Specifically, we have provided an example of how the model can calculate changes in real
estate values when economic variables, such as employment and median household income, are shocked.
Such a stark employment reduction has significant implications for real estate market conditions, and
further emphasizes the need for businesses in contaminated areas to implement long-term resiliency
measures that attract employment and investment. Furthermore, federal, state, and local government
agencies can use an analysis such as this as a tool to gauge incentives to stimulate this return, so as to
expedite the recovery process for homeowners and businesses.
We emphasize the caveat that our results are intended to be illustrative rather than precise figures
for policy formulation at this time. While the predictive model is statistically sound, its application could
still be improved. This would entail obtaining additional data on the distribution of original mortgage
levels and pre-attack equity levels in each of the three geographic zones in Seattle. The analysis could be
further bolstered by finding ways to differentiate the prediction of median home values by impact zone.
The final step would be to improve the supplementary analysis of the likelihood of default/foreclosure in
20
the event that home values drop significantly below mortgage balance levels. Unfortunately, there are
few real world examples upon which to base this analysis.
Endnotes vi Standard deviation estimates determined from 2008 Seattle home sales figures.
vii
Completed survey responses for both key questions provided a sample size (N) of 91 respondents.
viii
These estimations are indicative of how many housing units will be impacted by a foreclosed property. Our
definition of a housing unit includes single family (1 living unit), multi-family (2-3 living units), individual
condominium units, and individual apartment units. Therefore, when we say that 8,000 units will be foreclosed, we
assume that several foreclosed condominium and/or apartment complexes capture a significant amount of those
units. Our results inform us not of how many properties will be foreclosed, but how many individual housing units
will be negatively impacted by a foreclosed property.
21
References
Ahrens, Frank. "Enron's Just One Piece of Houston's Problems." The Seattle Times. Web. 22 January
2011. <http://community.seattletimes.nwsource.com/archive/?date=20020224&slug=houston24>.
American Association of Port Authorities. (2009). “North American Port Container Traffic Spreadsheet.”
Bureau of Economic Analysis. (2011). Personal Income for Metropolitan Areas, 2009.
City of Seattle. (2007). The Greater Seattle Datasheet. Retrieved on 08 August 2010 from
http://www.cityofseattle.net/oir/datasheet/economy.htm
Conway, Dick. (2010). “Puget Sound Economic Outlook Presentation,” The Puget Sound Economic
Forecaster. Seattle 38th Annual Economic Forecast Conference. January 14, 2010.
Conway, D., and Pedersen, D. (2010). “Are We There Yet?” The Puget Sound Economic Forecaster, 18,
1: March 2010.
Conway, D. and Pedersen, D. (2010). “A Difficult Recovery.” The Puget Sound Economic Forecaster,
18, 2: June 2010.
Conway, D. and Pedersen, D. (2011). “The Hold-Up.” The Puget Sound Economic Forecaster, 19, 2:
June 2011.
Conway Pedersen Economics, Inc. (2010). Current Economic Indicators. The Puget Sound Economic
Forecaster.
Conway Pedersen Economics, Inc. (2009). Puget Sound Economy Charts. The Puget Sound Economic
Forecaster
Schweitzer, Tamara. (2010). “How to Start a Business in Seattle.” Inc. July 26, 2010. Retrieved on 07
August 2010 from http://www.inc.com/guides/2010/07/how-to-start-a-business-in-seattle.html
Smith, V. K. et al. (2006). “Adjusting to Natural Disasters,” Journal of Risk and Uncertainty 33: 37-54.
United States Census Bureau. (2010). Population Estimates.
U.S. Census Bureau OnTheMap. (2011). Retrieved from http://lehdmap.did.census.gov/
University of Washington. (2010). The Economic and Societal Impact of the University of Washington.
July 7, 2010.
Wial, H. and Shearer, R.. (2010). Tracking Economic Recession and Recovery in America’s 100 Largest
Metropolitan Areas. MetroMonitor. Brookings Institution: Metropolitan Policy Program. June
2010.
Woodyard, C., and M. Kasindorf. "Troubles Ripple through Houston." USA Today. 21 January 2002.
1
CHAPTER 4. A PRO-FORMA INCOME STATEMENT MODEL TO ESTIMATE THE IMPACTS OF DISASTERS ON BUSINESS RATES OF RETURN
by
Morgan Bender and Gregory Blount
I. INTRODUCTION
This chapter presents a simulation model that can predict the impact of an anthrax attack and other
disasters on the profitability of individual businesses. The model is based on the standard Pro-Forma
Income Statement analysis commonly used in business valuation. It utilizes a firm’s financial statements
from the most recent two years of operation in order to project a baseline revenue growth rate and a
percentage of sales approach to project baseline expenses. Economic conditions from the anthrax attack,
such as the extent and duration of the regional economic downturn, are entered into the model to estimate
the negative impacts on the firm’s revenue, and the model proceeds to project the impact on the firm’s
bottom line.
In order to use this model for private businesses without readily available public financial statements,
industry averages can be used for the key parameters. The model can accommodate any changes in
economic conditions from the disaster.
The model is presented in a spreadsheet format to facilitate its use. Microsoft Excel is the platform.
II. MODEL DESCRIPTION The simulation model is based on the standard Pro-Forma Income Statement analysis. The model uses a
business’s financial statements from the most recent two years of operation as a benchmark to project a
baseline growth rate for the post-disaster period and a percentage of sales approach to project future
expenses. These are basic income statement entries.
.
The model predicts impacts on the rate of return stemming from changes in business conditions caused by
a disaster. Based on the likely affects from a terrorist attack, such as the release of anthrax, the key
variables adjusted in this the following analysis include:
● Projected Decline in Demand
● Duration of Demand Decline
● Duration of Facility Closing
● Decontamination Costs
● Operating Leverage
Other variables included in the analysis are:
● Baseline Growth Rate
● Pre-Disaster Assets
● Pre-Disaster Equity
● Facility Size
● Tax Rate
2
The resulting outputs from the model include:
Net Income (NI)
Return on Assets (ROA)
Return on Equity (ROE)
III. HOW THE MODEL WORKS
A. Model Set-Up
User determined firm characteristics:
Baseline Growth Rate: The expected baseline growth rate is used to forecast the expected sales for the
next five financial reporting periods. It is calculated using assumption based on a combination of the
company’s past two years of performance and on prevalent industry growth rates.
Pre-Disaster Assets and Equity: These variables form the basis of the calculation of both the disaster
and baseline ROA and ROE. Both variables can be found on the 10-K and 10-Q Forms the company has
filed with the Securities and Exchange Commission (SEC).
There are two key assumptions made in this forecast, the company issues no dividends and has no capital
expenditures during its recovery (reasonable assumptions for a company in the aftermath of a disaster,
since it is likely to attempt to build a security cushion during this volatile time). Thus, the company’s net
income in its entirety will be recorded as retained earnings and added to the owners’ equity line of the
balance sheet. For simplicity and based on the lack of capital expenditures, assets grow at the same rate
as equity.
Facility Size: The size of the facility is used in conjunction with the cost of decontamination to determine
the extraordinary item cost in year one of the projections.
Tax Rate: While companies are taxed on a marginal basis in a given year, companies have the ability to
forgo taxes and accumulate taxes payable over time. It is therefore the case that income tax expense often
deviates from the level of taxes a firm should pay based on its marginal tax bracket in a given year. The
tax rate should be assumed using an equally weighted combination of the firm’s effective and marginal
tax rates.
Post-Disaster changes and impacts:
Projected Decline in Demand: Following an anthrax attack, demand would decline for a number of
reasons including lack of consumer confidence and reduction in viable clientele do to mass flights from
the contaminated area. The model accounts for this decline in a simple manner by decreasing the
assumed normal sales projection by the proportion that the demand is assumed to decrease.
Duration of Demand Decline: This variable is necessary since an exact remediation time frame is
impossible to predict, too many variables are present (such as how many agencies can offer aid) and the
overall size of the area impacted. This variable, given in years, is applied to the model by reducing sales
growth in the applicable years.
Duration of Facility Closing: In many circumstances a firm will have to shut down completely for a
short time after the attack because of quarantine and/or decontamination. To account for this, the user can
3
input an assumed shutdown time measured in days. This is accounted for by decreasing the first year’s
revenue by the proportion of the year that the firm is shut down.
Decontamination Costs: Decontamination of sights where anthrax is present is a necessary function in
order to continue business operations. Therefore, this variable is used to calculate the extraordinary
cleanup expense in year one.
For this model a $5.70 per square foot figure will be utilized as a lower bound based on research done by
Canter et al. (2005). However, cleanup costs are highly variable as seen through historic cleanup costs
garnered from the Capitol Hill anthrax attack in 2001 (Price, 2009). Taking this into consideration,
decontamination costs are a variable input in the model and explicitly adjusted in Scenario 7.
Operating Leverage: The user can enter a likely estimated operating leverage, which is the proportion of
fixed costs relative to total costs. A low degree of operating leverage will allow a firm to minimize costs
as production slows down. A high degree of operating leverage will be dangerous for the firm because it
will still incur high fixed costs while its production wanes.
B. Explanation of Key Outputs
Net Income (NI): This shows the bottom line profitability of the company. The Net Income of a
company after the disaster can be compared to the pre-disaster Net Income or “but-for” Net Income (the
Net Income that the company believes it would have realized but for the disaster) for business insurance
purposes.
Return on Assets (ROA): One of several measures of operating efficiency, ROA measures the firm’s
return on total assets, in terms of net income as a percentage of total assets. The significance of this
measure comes from comparing operating efficiency to other similar companies. A company that earns a
higher return on assets is assumed to be more efficient, as it is able to earn more per given amount of
assets. Investors and lenders often look at this measure in order to evaluate a company’s worthiness of
new investment. A company with a relatively low ROA will be viewed as less worthy, since it is
assumed that it should be able to increase its earnings though streamlining operations and increasing
ROA.
Return on Equity (ROE): Return on Equity measures the firm’s returns on shareholders equity,
calculated as Net Income as a percentage of shareholders equity. ROE is arguably the most important
measure to investors. At the very least, it is the first thing looked at when screening companies. Investors
look at this ratio in a similar manner to ROA; however, this is more important because it represents the
return, or profit, that a company is generating on the money contributed by investors. Again, similar to
ROA, this ratio is viewed relatively to other similar companies. It is usually correlated with risk, as
shown in the Capital Asset Pricing Model, which models the expected return based on a company’s
riskiness (see below).
Implied Beta: In investing, beta represents the “riskiness” of a company. It is measured as the variance
of the stock price of a company with respect to the variance of the S&P 500, where the beta of the market
is 1.0. A Beta higher than 1 implies that the when the market average swings up or down, the stock of
this company will do so by a greater degree; a beta value lower than 1 implies the opposite. Based on this
logic, investors in companies that have a beta greater than 1, thus classifying it as more risky than
investing in a market index, will expect a higher rate of return than the market average.
Our model is applied to companies experiencing an extreme shock. Although it would be difficult and
subjective to attempt to calculate a new beta for the company in the post-disaster period, it may be useful
to evaluate the implied beta based on the new Return on Equity ratios. This will provide some insight as
4
to the level of risk an investor will be willing to take on for the given return on equity. As can be seen in
the model, the early post-disaster ROE implies that investors would expect the riskiness of the investment
to be significantly below that of investing in the market index. Since it is obvious that a company will be
in a much more vulnerable state at this time, we can infer that the beta is, in reality, higher than that of the
historical average prior to the disaster. This means that there will be considerable degree of disinvestment
in the said firm.
Table 1 defines each variable and parameter used in the model, as well as the prime source of data. The
table also lists some key assumptions in its final column.
IV. EXPLANATION OF KEY CALCULATIONS
The calculations that follow are derived in the appendices of each scenario.
Sales: To calculate sales a combination of the variables Baseline Growth Rate, Projected Decline in
Demand, Duration of Demand Decline, and Duration of Facility Closing were used. Sales increase with
growth and are reduced in the first year by the percentage that demand declines and by the percentage of
one year that the facility remains closed. In the forward projections demand jumps up by the percentage
of decline in demand following the last year effected.
Cost of Goods Sold and Selling General and Admin Expense: Both of these expenses are calculated
using their levels and growth rates from the previous two years. A weighted average of the variable’s
ratios to sales was utilized to project them forward as a percentage of sales.
Depreciation, Depletion, and Amortization: Several assumptions were made in regards to this variable
in light of complexities and differences in recording procedures within Generally Accepted Accounting
Principles (GAAP). It is assumed that the companies use straight line depreciation (assets depreciate or
amortize at a constant rate each year) and no asset fully depreciates in the 5 year projection. As per these
assumptions and the assumption previously made in Section III in regards to zero capital expenditures
during the firm’s recovery phase this variable remains constant.
Interest Expense: Some complexity arises when discussing this variable. As firms contract after an
anthrax attack, it is logical to assume that they would seek additional debt (within their leverage bounds
and their ability to pay an increased interest expense) to forgo bankruptcy. It would however take another
extensive study to fully comprehend how the debt markets would react to such an attack and to the plight
of Seattle based firms. What’s more, there is a likely hood of government intervention that would have to
be taken into consideration. Given these complexities it is assumed that the firms refrain from taking on
additional debt or paying down principal, and that no debt matures in the 5-year projection. Given these
assumptions interest expense remains constant.
Minority Interest in Income: When a firm controls 50% or more of another company, the controlling
firm must consolidate its balance sheet with the controlled company’s balance sheet. Because the income
statements are combined without 100% control, a portion of the reported income must go to the minority
interest. Just as it is assumed that no capital expenditures are made during the recovery phase, the model
assumes that no acquisitions are undertaken. Minority interest therefore grows as a percentage of sales
based on a weighted average rate calculated from the previous two business years.
Non-Operating Income: This is a highly volatile line item that has a myriad of internal and external
events that trigger its increase or decrease. Because of its unknown nature the previous business year’s
non-operating income is taken as a normalized level and is used consistently in the 5-year projection.
5
CH
AR
T 1
. M
OD
EL
VA
RIA
BL
ES,
PA
RA
ME
TE
RS,
AN
D O
UT
PUT
S
Var
iabl
eD
efin
itio
n U
nit
Dat
a So
urce
Com
men
ts
Bas
elin
e G
row
th R
ate
Ass
umed
futu
re g
row
th r
ate
of s
ales
in n
orm
al
busi
ness
clim
ate
%A
ssum
ptio
n ba
sed
on p
ast 2
yea
rs p
erfo
rman
ce.
Indu
stry
gro
wth
rat
e
Pre-
Dis
aste
r Sa
les
Sale
s fo
r th
e 12
mon
ths
prio
r to
dis
aste
r$K
Com
pany
file
d 10
-K r
epor
ts (
for
pre-
disa
ster
per
iods
)
Pre-
Dis
aste
r (2
010)
Ass
ets
Val
ue o
f tot
al a
sset
s in
yea
r pr
ior
to d
isas
ter
$KC
ompa
ny fi
led
10-K
rep
orts
(fo
r pr
e-di
sast
er p
erio
ds)
Pre-
Dis
aste
r (2
010)
Equ
ityV
alue
of o
wne
rs e
quity
in y
ear
prio
r to
dis
aste
r$K
Com
pany
file
d 10
-K r
epor
ts (
for
pre-
disa
ster
per
iods
)
Ope
ratin
g L
ever
age
Perc
enta
ge o
f fix
ed c
osts
rel
ativ
e to
tota
l cos
ts%
Cal
cula
ted
as fi
xed
cost
s re
lativ
e to
tota
l cos
ts o
r as
an
indu
stry
aver
age.
Indu
stry
ave
rage
Proj
ecte
d D
eclin
e in
Dem
and
Proj
ecte
d de
clin
e in
dem
and
as a
res
ult o
f the
disa
ster
%A
ssum
ptio
nE
xter
nal s
hock
Dur
atio
n of
Dem
and
Dec
line
Proj
ecte
d du
ratio
n of
the
decl
ine
in d
eman
d fo
r
prod
uct
Yea
rsA
ssum
ptio
nE
xter
nal s
hock
Dur
atio
n of
Fac
ility
Clo
sing
Proj
ecte
d du
ratio
n of
faci
lity
clos
ure
as a
res
ult o
f
disa
ster
Day
sA
ssum
ptio
nE
xter
nal s
hock
Dec
onta
min
atio
n C
osts
Proj
ecte
d de
cont
amin
atio
n co
st
Sq. F
eet
Res
earc
h by
Can
ter
et a
l. (2
005)
and
Blo
unt (
2011
)
Faci
lity
Size
Size
of t
he c
onta
min
ated
faci
lity
Sq. F
eet
Com
pany
file
d 10
-K r
epor
ts (
for
pre-
disa
ster
per
iods
)
Tax
Rat
ePo
rtio
n of
inco
me
paid
to th
e go
vern
men
t %
Ass
umpt
ion
base
d on
pas
t 2 y
ears
.
Inco
me
Stat
emen
t Ite
ms
Indi
vidu
al r
even
ues
and
expe
nditu
res
$KC
ompa
ny fi
led
10-K
rep
orts
(fo
r pr
e-di
sast
er p
erio
ds)
and
base
d on
perc
enta
ge o
f sal
es (
for
fore
cast
ed v
alue
s)
Ret
urn
on A
sset
sR
ate
of r
etur
n on
the
com
pany
's as
sets
, mea
sure
d as
net i
ncom
e re
lativ
e to
ass
ets
%C
alcu
late
d fr
om C
ompa
ny 1
0-K
(fo
r pr
e-di
sast
er p
erio
ds)
and
from
fore
cast
(fo
r po
st-d
isas
ter
peri
ods)
Ass
umpt
ion:
Pri
or y
ear's
net
inco
me
is
adde
d to
ass
ets
Ret
urn
on E
quity
Rat
e of
ret
urn
on th
e co
mpa
ny's
equi
ty, m
easu
red
as
net i
ncom
e re
lativ
e to
ow
ners
equ
ity%
Cal
cula
ted
from
Com
pany
10-
K (
for
pre-
disa
ster
per
iods
) an
d fr
om
fore
cast
(fo
r po
st-d
isas
ter
peri
ods)
Ass
umpt
ion:
Pri
or y
ear's
net
inco
me
is
adde
d to
ret
aine
d ea
rnin
gs
Ret
urn
on E
quity
: Mod
el;
Mar
ket R
etur
n: A
vg. r
etur
n on
S&
P 50
0 si
nce
1950
;
Ris
k Fr
ee r
ate:
10
year
Tre
asur
y Y
ield
Impl
ied
Bet
a
Impl
ied
Bet
a va
lue
base
d on
the
Ret
urn
on E
quity
,
risk
free
rat
e, r
ate
of r
etur
n on
the
mar
ket a
nd th
e
Cap
ital A
sset
Pri
cing
Mod
el e
quat
ion.
#
6
Baseline Net Income: This figure is used to compare post-disaster net income to normalized projections.
In order to calculate this normalized income a second pro-forma income statement (Baseline Income
Statement) was created following a Baseline Growth Rate that is unaffected by the disaster variables:
Projected Decline in Demand, Duration of Decline in Demand, and Duration of Facility Closing.
Baseline ROA and ROE: These figures are calculated in the same way as Post-Disaster ROA and ROE,
except they derive their growth from the Baseline Income Statement.
Extraordinary Item, Year One: This figure is calculated by multiplying facility size by the
decontamination cost per square foot.
V. IDENTIFYING APPROPRIATE TARGET COMPANIES
The portion of Seattle presumed to be attacked with anthrax contains a number of different firms.
However, two problems arise in finding appropriate target companies. The first is that many of those
firms are private, and the SEC has ruled that privately held businesses are not required to release public
financial statements; because of this ruling it is nearly impossible to gain access to information for many
of the businesses that would be affected in the Seattle area. The second problem is that public firms
operate, more often than not, on a national or global scale. As the model focuses on Seattle-based firms,
it can be difficult to discern what part of a public firm’s national or global business is derived from the
Seattle area.
A. Solutions for Private Companies
In light of a firm’s right to keep secret its financial statements, there are two different ways that
academically sound projections can be made for private firms.
Industry Averages: Sales figures are often more readily available than any other financial information
for private firms, via news releases and other media driven communications. Once the sales figure is in
hand a compilation of industry averages can be used to estimate costs as a percentage of sales, as well as
other variables such as assets and equity. Using this approach will give realistic projections as most firms
hover around their respected industry averages.
Recent Initial Public Offering (IPO): When firms go public they release an S-1 filing to the SEC. This
filing is geared towards creating confidence in investors and therefore, typically incorporates consolidated
financial data for the trailing 2 – 3 years. This consolidated data gives a glimpse into their operations as a
private firm and can be used for projections as long as the IPO was relatively recent. The user of this
approach should be cautious in the selection process, as many private businesses that go public do so
because they are in a state of negative earnings. They seek an IPO to boost production, advertisement,
etc. in order to become profitable. A firm with negative profitability may give skewed results.
B. Solution for National or Global Public Firms
The ideal public firm will separate its Seattle operations and revenues from its non-Seattle operations.
With this data a user will be able to adjust the input numbers so that the effects of an anthrax attack
represent those on its operations in the Seattle area alone. To do this, it is appropriate to scale down the
values used in the model by the appropriate size to represent Seattle area operations. The simplest way to
do this is to determine what percentage of total company revenues is contributed by the Seattle branch.
This ratio can be used on all income statement items and even on balance sheet items such as assets and
stockholders’ equity.
7
It is important to realize that the output numbers from the model when adjusted in this manner only
represent the Seattle area, but will not be viewed in this manner by investors. In order to calculate the
overall effects to the public institution, the model must be run again, under completely normal conditions
with income statement numbers representing non-Seattle revenue ratios. This will calculate the normal
expected operating profit for the non-affected areas so that the Seattle values can be factored in. This will
enable the user of the model to calculate the overall expected post-disaster net income, ROA, ROE and
implied beta.
VI. SAMPLE RESULTS
A. Selected Companies
Esterline Technologies: This public company was founded in 1978 and is a specialized manufacturing
firm with principally serving defense and aerospace markets. It runs three different segments on a global
scale; those segments are Avionics & Controls, Sensors & Systems, and Advanced Materials.
Esterline was chosen because it is a Seattle based firm that goes into great detail as to how its revenues
are divided amongst its multiple manufacturing plants. This made it relatively easy to separate out its
Seattle manufacturing plant and consequently scale down the necessary input variables.
Symetra Financial: This public company was founded more than 50 years ago and provides retirements,
benefit and life insurance products. It has a singular location in the Seattle area.
Symetra Financial was chosen because it is a Seattle based firm and went public in 2010. Through the S-
1 form it filed with the SEC, its private company financials for the years 2008 and 2009 were available
for use in analysis.
B. Understanding and Utilizing the Scenarios
All together there are eight scenarios, four for each company. Each scenario has been built up from the
lower bound, which uses the following disaster based inputs:
● Projected Decline in Demand: 20%
● Duration of Demand Decline: 3 years
● Duration of Facility Closing: 60 days
● Decontamination Costs: $5.70 per square foot
The parameters that change from scenario to scenario have been italicized in the tables of simulation
results below so that they may be more readily identifiable.
8
C. Discussion of Results
TABLE 1
Table 1 presents the model applied to the Seattle based firm Esterline Technologies. It simulates the
effects of an anthrax attack on Net Income, ROA, and ROE using a projected growth rate in sales of 12%
and operating leverage of 20%. Following projections for the lower bound, if demand declines by 20%
for three years, the output parameters for the company decline but remain positive for all three years.
Even though in year one NI, ROA, and ROE decrease by 59%, 5.6%, and 10.3% respectively, the
company would likely stay in operation. Two important elements that should be noted are the effects of
reopening the facility realized in year two, and the expiration of the decline in demand realized in year 4
for this scenario. These two events lead to positive jumps in sales for both years two and four. This is a
commonality amongst the different scenarios. It is also important to note that a company in a weaker
financial position relative to Esterline may experience a greater impact from the decline in demand, which
may cause negative returns in the first few years. This concept is shown in the analysis of Symetra
Financial and is presented in Table 4.
Scenario 1: Esterline TechnologiesLower Bound
Baseline Growth Rate (%) 12%
Pre-Disaster Assets (2010) 517,547
Pre-Disaster Equity (2010) 282,559
Operating Leverage (%) 20%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 60
Facility Size (square feet) 216,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2009 2010 2011E 2012E 2013E 2014E 2015E
Sales (Net) 281,492 305,320 228,597 298,115 333,889 448,747 502,597
Cost of Goods Sold 190,832 202,078 153,136 199,706 223,670 300,613 336,686
Gross Profit 90,660 103,242 75,461 98,410 110,219 148,134 165,910
Selling, General, & Admin Expenses 33,264 37,730 29,651 36,373 39,832 50,939 56,146
Depreciation, Depletion, & Amortization 13,833 13,928 13,928 13,928 13,928 13,928 13,928
Operating Income 43,563 51,584 31,882 48,109 56,459 83,267 95,836
Interest Expense 5,738 6,636 6,636 6,636 6,636 6,636 6,636
Minority Interest in Income 327 192 205 267 299 402 450
Non-Operating Income (Expenses) 1,594 2 2 2 2 2 2
Pretax Income 35,904 44,758 25,040 41,204 49,522 76,228 88,748
Income Taxes 2,510 4,901 3,756 6,181 7,428 11,434 13,312
Income Before Extraordinary Items & Discontinued
Operations 33,394 39,857 21,284 35,023 42,093 64,793 75,436
Extraordinary Items and Discontinued Operations (net of
income taxes) 0 0 1,231 0 0 0 0
Post-Disaster Net Income 33,394 39,857 20,053 35,023 42,093 64,793 75,436
Baseline Net Income 33,394 39,857 49,133 57,916 67,752 78,770 91,109
Difference -59% -40% -38% -18% -17%
Post Disaster Return on Assets (ROA) 7.7% 3.9% 6.3% 7.1% 9.8% 10.3%
Baseline Return on Assets (ROA) 7.7% 9.5% 10.1% 10.5% 10.9% 11.2%
Difference (baseline projection vs. post-disaster) 0.0% -5.6% -3.7% -3.5% -1.1% -0.9%
Post Disaster Return on Equity (ROE) 14.1% 7.1% 11.0% 11.7% 15.3% 15.1%
Baseline Return on Equity (ROE) 14.1% 17.4% 17.0% 17.0% 16.9% 16.7%
Difference (baseline projection vs. post-disaster) 0.0% -10.3% -6.0% -5.3% -1.6% -1.6%
Implied Beta 1.41 0.48 1.00 1.09 1.57 1.55
Va
ria
ble
s a
nd
Pa
ram
eter
s
9
TABLE 2
Table 2 presents results for a scenario based on parameters that are in-line with our upper bound
projections: projected demand decline of 80% for five years. The results of the analysis demonstrate the
devastating effects that an anthrax attack could have. Projected NI, ROA, and ROE fall by 133%, 12.6%,
and 23.1% respectively, more than double the decrease in Scenario 1, and are all negative. These
negative rates are present in each of the five projected years. An interesting result of this scenario is that
NI, ROE, and ROA recover at a slower rate than in Scenario 1.
NI recovers by 42% in Scenario 1 and by 29% in Scenario 2.
ROA recovers by 4.5% in Scenario 1 and by .6% in Scenario 2.
ROE recovers by 8.7% in Scenario 2 and by 4.9% in Scenario 2.
In this case, the anthrax attack is devastating, resulting in negative rates of return for the following four
years. Net income loss for the firm is projected to be $16 million in 2011. An average outcome like this
for just 100 firms in Seattle following the attack would represent losses of $1.6 billion. There are of
course many more firms operating Seattle, but most of the larger ones are part of multi-plant firms that
can better withstand the disaster by shifting economic activity elsewhere. This means that they will
require less government assistance themselves, but this shift in economic still will represent a major loss
to Seattle. It emphasizes the need for not just subsidies to help local businesses survive but also actions
that will promote recovery of all firms in the Seattle area.
Scenario 2: Esterline TechnologiesUpper Bound
Baseline Growth Rate (%) 12%
Pre-Disaster Assets (2010) 517,547
Pre-Disaster Equity (2010) 282,559
Operating Leverage (%) 20%
Projected Decline in Demand (%) 80%
Duration of Demand Decline (years) 5
Duration of Facility Closing (days) 60
Facility Size (square feet) 216,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2009 2010 2011E 2012E 2013E 2014E 2015E
Sales (Net) 281,492 305,320 57,149 74,529 83,472 93,489 104,708
Cost of Goods Sold 190,832 202,078 38,284 49,926 55,918 62,628 70,143
Gross Profit 90,660 103,242 18,865 24,602 27,555 30,861 34,565
Selling, General, & Admin Expenses 33,264 37,730 13,072 14,753 15,618 16,586 17,671
Depreciation, Depletion, & Amortization 13,833 13,928 13,928 13,928 13,928 13,928 13,928
Operating Income 43,563 51,584 8,135 4,078 1,991 347 2,966
Interest Expense 5,738 6,636 6,636 6,636 6,636 6,636 6,636
Minority Interest in Income 327 192 51 67 75 84 94
Non-Operating Income (Expenses) 1,594 2 2 2 2 2 2
Pretax Income 35,904 44,758 14,824 10,783 8,704 6,375 3,766
Income Taxes 2,510 4,901 0 0 0 0 0
Income Before Extraordinary Items & Discontinued
Operations 33,394 39,857 14,824 10,783 8,704 6,375 3,766
Extraordinary Items and Discontinued Operations (net of
income taxes) 0 0 1,231 0 0 0 0
Post-Disaster Net Income 33,394 39,857 16,055 10,783 8,704 6,375 3,766
Baseline Net Income 33,394 39,857 49,133 57,916 67,752 78,770 91,109
Difference -133% -119% -113% -108% -104%
Post Disaster Return on Assets (ROA) 7.7% -3.1% -2.1% -1.7% -1.3% -0.8%
Baseline Return on Assets (ROA) 7.7% 9.5% 10.1% 10.5% 10.9% 11.2%
Difference (baseline projection vs. post-disaster) 0.0% -12.6% -12.2% -12.3% -12.2% -12.0%
Post Disaster Return on Equity (ROE) 14.1% -5.7% -4.0% -3.3% -2.5% -1.5%
Baseline Return on Equity (ROE) 14.1% 17.4% 17.0% 17.0% 16.9% 16.7%
Difference (baseline projection vs. post-disaster) 0.0% -23.1% -21.0% -20.3% -19.4% -18.2%
Implied Beta 1.41 -1.22 -1.00 -0.91 -0.80 -0.67
Va
ria
ble
s a
nd
Pa
ram
eter
s
10
TABLE 3
Scenario 3, assumes Esterline Technologies has an operating leverage of 80% rather than 20%. While the
company’s returns remain positive for all five of the projected years they decline relative to Scenario 1.
The differences in NI falls to -69% from -59%, ROA falls to -6.5% from -5.6%, ROE falls to -12% from -
10.3% compared to baseline projections. These results agree with the discussion on operating leverage
found in the “User determined firm characteristics” section. Variable costs decrease with sales revenue,
but the increased portion of fixed costs remain stable leading to greater decreases in NI and thus ROA and
ROE relative to Scenario 1. This scenario demonstrates that companies with a high degree of operating
leverage, such as manufacturing firms, will have a harder time staying in business following an anthrax
attack than would service companies such as Symetra Financial.
Scenario 3: Esterline TechnologiesIncreased Operating Leverage
Baseline Growth Rate (%) 12%
Pre-Disaster Assets (2010) 517,547
Pre-Disaster Equity (2010) 282,559
Operating Leverage (%) 80%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 60
Facility Size (square feet) 216,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2009 2010 2011E 2012E 2013E 2014E 2015E
Sales (Net) 281,492 305,320 228,597 298,115 333,889 448,747 502,597
Cost of Goods Sold 190,832 202,078 153,136 199,706 223,670 300,613 336,686
Gross Profit 90,660 103,242 75,461 98,410 110,219 148,134 165,910
Selling, General, & Admin Expenses 33,264 37,730 35,710 37,391 38,256 41,032 42,334
Depreciation, Depletion, & Amortization 13,833 13,928 13,928 13,928 13,928 13,928 13,928
Operating Income 43,563 51,584 25,823 47,091 58,035 93,174 109,648
Interest Expense 5,738 6,636 6,636 6,636 6,636 6,636 6,636
Minority Interest in Income 327 192 205 267 299 402 450
Non-Operating Income (Expenses) 1,594 2 2 2 2 2 2
Pretax Income 35,904 44,758 18,980 40,186 51,098 86,134 102,560
Income Taxes 2,510 4,901 2,847 6,028 7,665 12,920 15,384
Income Before Extraordinary Items & Discontinued
Operations 33,394 39,857 16,133 34,158 43,434 73,214 87,176
Extraordinary Items and Discontinued Operations (net of
income taxes) 0 0 1,231 0 0 0 0
Post-Disaster Net Income 33,394 39,857 14,902 34,158 43,434 73,214 87,176
Baseline Net Income 33,394 39,857 48,712 59,753 72,120 85,970 101,483
Difference -69% -43% -40% -15% -14%
Post Disaster Return on Assets (ROA) 7.7% 2.9% 6.2% 7.3% 11.0% 11.5%
Baseline Return on Assets (ROA) 7.7% 9.4% 10.4% 11.1% 11.7% 12.1%
Difference (baseline projection vs. post-disaster) 0.0% -6.5% -4.2% -3.8% -0.7% -0.6%
Post Disaster Return on Equity (ROE) 14.1% 5.3% 10.8% 12.1% 16.9% 16.7%
Baseline Return on Equity (ROE) 14.1% 17.2% 17.5% 17.9% 18.1% 18.1%
Difference (baseline projection vs. post-disaster) 0.0% -12.0% -6.7% -5.9% -1.2% -1.4%
Implied Beta 1.41 0.24 0.97 1.14 1.79 1.77
Va
ria
ble
s a
nd
Pa
ram
eter
s
11
TABLE 4
Scenario 4 simulates the effects of a cleanup time that lasts 120 days rather than 30 days. Given the size
of the area contaminated with anthrax and the difficulty of cleaning up an exposed area, it is reasonable to
assume that facilities may be closed for a duration that is longer than 30 days (Price, 2009). Year one
returns are significantly lower with facility closure set to 120 days rather than 30 days, with the
differences in NI falling to -77% from -59%, ROA falling to -7.3% from -5.6%, and ROE falling to
-13.4% from -10.3%. The key outputs are more negatively impacted by an increase in facility closing
days than they are from an increase in operating leverage, as per Scenario 3. This scenario demonstrates
the critical nature of the remediation time frame and how important it is to have a well formulated and
efficient anthrax attack response plan.
Scenario 4: Esterline TechnologiesIncreased Facility Closing Days
Baseline Growth Rate (%) 12%
Pre-Disaster Assets (2010) 517,547
Pre-Disaster Equity (2010) 282,559
Operating Leverage (%) 20%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 120
Facility Size (square feet) 216,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2009 2010 2011E 2012E 2013E 2014E 2015E
Sales (Net) 281,492 305,320 183,627 273,277 306,070 411,359 460,722
Cost of Goods Sold 190,832 202,078 123,011 183,067 205,035 275,566 308,634
Gross Profit 90,660 103,242 60,616 90,211 101,036 135,792 152,087
Selling, General, & Admin Expenses 33,264 37,730 25,302 33,971 37,142 47,324 52,097
Depreciation, Depletion, & Amortization 13,833 13,928 13,928 13,928 13,928 13,928 13,928
Operating Income 43,563 51,584 21,386 42,311 49,965 74,541 86,062
Interest Expense 5,738 6,636 6,636 6,636 6,636 6,636 6,636
Minority Interest in Income 327 192 164 245 274 368 412
Non-Operating Income (Expenses) 1,594 2 2 2 2 2 2
Pretax Income 35,904 44,758 14,584 35,429 43,053 67,534 79,012
Income Taxes 2,510 4,901 2,188 5,314 6,458 10,130 11,852
Income Before Extraordinary Items & Discontinued
Operations 33,394 39,857 12,396 30,114 36,595 57,404 67,160
Extraordinary Items and Discontinued Operations (net of
income taxes) 0 0 1,231 0 0 0 0
Post-Disaster Net Income 33,394 39,857 11,165 30,114 36,595 57,404 67,160
Baseline Net Income 33,394 39,857 49,133 57,916 67,752 78,770 91,109
Difference -77% -48% -46% -27% -26%
Post Disaster Return on Assets (ROA) 7.7% 2.2% 5.5% 6.3% 8.9% 9.5%
Baseline Return on Assets (ROA) 7.7% 9.5% 10.1% 10.5% 10.9% 11.2%
Difference (baseline projection vs. post-disaster) 0.0% -7.3% -4.6% -4.3% -2.0% -1.7%
Post Disaster Return on Equity (ROE) 14.1% 4.0% 9.6% 10.5% 14.1% 14.2%
Baseline Return on Equity (ROE) 14.1% 17.4% 17.0% 17.0% 16.9% 16.7%
Difference (baseline projection vs. post-disaster) 0.0% -13.4% -7.4% -6.5% -2.8% -2.5%
Implied Beta 1.41 0.06 0.82 0.93 1.42 1.42
Va
ria
ble
s a
nd
Pa
ram
eter
s
12
TABLE 5
Scenario 5 examines the results for Symetra Financial given lower bound parameter estimates. NI in the
first year falls by 88%, nearly 30% more than the drop Esterline Technologies was shown to experience in
Scenario 1. This heightened fall can be explained by both the smaller size of Symetra and its lower
margins relative to Esterline Technologies. As is the case with all locales, there are a greater number of
smaller firms in Seattle that lack the benefits of economies of scale and the proprietary knowledge that
larger firms like Esterline Technologies possess. These firms would experience the largest burdens
following an anthrax attack and thus have the largest probability of fiscal demise.
Scenario 5: Symetra FinancialLower Bound
Baseline Growth Rate (%) 8%
Pre-Disaster Assets (2009) 22,226
Pre-Disaster Equity (2009) 1,481
Operating Leverage (%) 5%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 60
Facility Size (square feet) 5,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2008 2009 2010E 2011E 2012E 2013E 2014E
Sales (Net) 1,451.10 2,245.63 1,621 2,039 2,202 2,854 3,082
Cost of Goods Sold 1,114.60 1,568.69 1,189 1,495 1,615 2,093 2,260
Gross Profit 336.50 676.94 432 544 587 761 822
Selling, General, & Admin Expenses 265.80 328.59 270 336 361 463 499
Depreciation, Depletion, & Amortization 25.80 63.70 64 64 64 64 64
Operating Income 44.90 284.65 98 144 162 234 259
Interest Expense 31.90 31.90 32 32 32 32 32
Minority Interest in Income 0.00 0.00 0 0 0 0 0
Non-Operating Income (Expenses) 0.00 0.00 0 0 0 0 0
Pretax Income 13.00 252.75 67 113 131 202 227
Income Taxes 9.10 68.95 10 17 20 30 34
Income Before Extraordinary Items & Discontinued
Operations 22.10 183.80 57 96 111 172 193
Extraordinary Items and Discontinued Operations (net of
income taxes) 0.00 0.00 29 0 0 0 0
Post-Disaster Net Income 22.10 183.80 28 96 111 172 193
Baseline Net Income 22.10 183.80 227 253 281 311 344
Difference -88% -62% -61% -45% -44%
Post Disaster Return on Assets (ROA) 0.8% 0.1% 0.4% 0.5% 0.8% 0.8%
Baseline Return on Assets (ROA) 0.8% 1.0% 1.1% 1.2% 1.3% 1.5%
Difference (baseline projection vs. post-disaster) 0.0% -0.9% -0.7% -0.7% -0.6% -0.6%
Post Disaster Return on Equity (ROE) 12.4% 1.9% 6.1% 6.6% 9.3% 9.4%
Baseline Return on Equity (ROE) 12.4% 15.4% 14.6% 14.1% 13.7% 13.3%
Difference (baseline projection vs. post-disaster) 0.0% -13.5% -8.5% -7.6% -4.5% -3.9%
Implied Beta 1.19 -0.21 0.34 0.41 0.77 0.79
Va
ria
ble
s a
nd
Pa
ram
eter
s
13
TABLE 6
Table 6 presents a scenario that uses upper bound projections, decline in demand of 80% for five year,
applied to Symetra Financial. As is expected from both the results of Scenario 2 and Scenario 5,
Symetra’s returns fall into the negatives for all five of the projected years. The large Net Losses (NL)
Symetra Financial over the five years indicate bankruptcy for the firm. Unless a company has substantial
cash flows or a large cash reserve that could be used to meet necessary payments for the duration of the
decline in demand, that company would be forced to declare bankruptcy. It is more likely that smaller
firms lack large reserves of cash and so most would close down. While large companies may continue
operations, the largest portion of Seattle based businesses would fail if demand declined by 80% for five
years. An anthrax attack of this magnitude would be catastrophic for the local economy and could total
the business sector; it would no longer be fiscally viable to clean the city. A high level of bankruptcy
proceedings could further disrupt the economic conditions of Seattle and lead to increased hardship for
companies that have the fiscal strength to withstand the decline in demand.
Scenario 6: Symetra FinancialUpper Bound
Baseline Growth Rate (%) 8%
Pre-Disaster Assets (2009) 22,226
Pre-Disaster Equity (2009) 1,481
Operating Leverage (%) 5%
Projected Decline in Demand (%) 80%
Duration of Demand Decline (years) 5
Duration of Facility Closing (days) 60
Facility Size (square feet) 5,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2008 2009 2010E 2011E 2012E 2013E 2014E
Sales (Net) 1,451.10 2,245.63 405 510 550 595 642
Cost of Goods Sold 1,114.60 1,568.69 297 374 404 436 471
Gross Profit 336.50 676.94 108 136 147 159 171
Selling, General, & Admin Expenses 265.80 328.59 80 96 103 109 117
Depreciation, Depletion, & Amortization 25.80 63.70 64 64 64 64 64
Operating Income 44.90 284.65 35 24 19 15 9
Interest Expense 31.90 31.90 32 32 32 32 32
Minority Interest in Income 0.00 0.00 0 0 0 0 0
Non-Operating Income (Expenses) 0.00 0.00 0 0 0 0 0
Pretax Income 13.00 252.75 67 56 51 47 41
Income Taxes 9.10 68.95 0 0 0 0 0
Income Before Extraordinary Items & Discontinued
Operations 22.10 183.80 67 56 51 47 41
Extraordinary Items and Discontinued Operations (net of
income taxes) 0.00 0.00 29 0 0 0 0
Post-Disaster Net Income 22.10 183.80 96 56 51 47 41
Baseline Net Income 22.10 183.80 227 253 281 311 344
Difference -142% -122% -118% -115% -112%
Post Disaster Return on Assets (ROA) 0.8% -0.4% -0.3% -0.2% -0.2% -0.2%
Baseline Return on Assets (ROA) 0.8% 1.0% 1.1% 1.2% 1.3% 1.5%
Difference (baseline projection vs. post-disaster) 0.0% -1.5% -1.4% -1.5% -1.6% -1.7%
Post Disaster Return on Equity (ROE) 12.4% -6.5% -3.9% -3.7% -3.5% -3.2%
Baseline Return on Equity (ROE) 12.4% 15.4% 14.6% 14.1% 13.7% 13.3%
Difference (baseline projection vs. post-disaster) 0.0% -21.8% -18.5% -17.9% -17.2% -16.5%
Implied Beta 1.19 -1.33 -0.99 -0.97 -0.93 -0.90
Va
ria
ble
s a
nd
Pa
ram
eter
s
14
TABLE 7
Scenario 7 examines the impacts of remediation rates of $57.00 per square foot, ten times higher than the
base case of $5.70 per square foot. NI in year one falls to -$228,000 from $28,000 in Scenario 5, due to
an increase in Extraordinary Items from $29,000 in Scenario 5 to $285,000. This high NL could push
firms into bankruptcy and force them to close. However, it is important to note that this negative impact
would only affect the firm in year one. Year two sales in Scenarios 5 and 7 are equivalent indicating that
projected NI following year one will be positive, given lower bound parameters. These positive returns
could justify staying in business even in the face of a large NL in year one. While a NI drop of $256,000
would have a smaller impact for larger firms with NIs in the millions, those firms may have larger
facilities. Larger facilities would magnify the impact of increased remediation rates causing cleanup costs
to be just as fiscally trying as they are for smaller firms.
Scenario 7: Symetra FinancialIncreased Cleanup Costs
Baseline Growth Rate (%) 8%
Pre-Disaster Assets (2009) 22,226
Pre-Disaster Equity (2009) 1,481
Operating Leverage (%) 5%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 60
Facility Size (square feet) 5,000
Decontamination Costs (per square foot) 57.00
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2008 2009 2010E 2011E 2012E 2013E 2014E
Sales (Net) 1,451.10 2,245.63 1,621 2,039 2,202 2,854 3,082
Cost of Goods Sold 1,114.60 1,568.69 1,189 1,495 1,615 2,093 2,260
Gross Profit 336.50 676.94 432 544 587 761 822
Selling, General, & Admin Expenses 265.80 328.59 270 336 361 463 499
Depreciation, Depletion, & Amortization 25.80 63.70 64 64 64 64 64
Operating Income 44.90 284.65 98 144 162 234 259
Interest Expense 31.90 31.90 32 32 32 32 32
Minority Interest in Income 0.00 0.00 0 0 0 0 0
Non-Operating Income (Expenses) 0.00 0.00 0 0 0 0 0
Pretax Income 13.00 252.75 67 113 131 202 227
Income Taxes 9.10 68.95 10 17 20 30 34
Income Before Extraordinary Items & Discontinued
Operations 22.10 183.80 57 96 111 172 193
Extraordinary Items and Discontinued Operations (net of
income taxes) 0.00 0.00 285 0 0 0 0
Post-Disaster Net Income 22.10 183.80 228 96 111 172 193
Baseline Net Income 22.10 183.80 227 253 281 311 344
Difference -200% -62% -61% -45% -44%
Post Disaster Return on Assets (ROA) 0.8% -1.0% 0.4% 0.5% 0.8% 0.8%
Baseline Return on Assets (ROA) 0.8% 1.0% 1.1% 1.2% 1.3% 1.5%
Difference (baseline projection vs. post-disaster) 0.0% -2.1% -0.7% -0.7% -0.6% -0.6%
Post Disaster Return on Equity (ROE) 12.4% -15.4% 6.1% 6.6% 9.3% 9.4%
Baseline Return on Equity (ROE) 12.4% 15.4% 14.6% 14.1% 13.7% 13.3%
Difference (baseline projection vs. post-disaster) 0.0% -30.8% -8.5% -7.6% -4.5% -3.9%
Implied Beta 1.19 -2.52 0.34 0.41 0.77 0.79
Va
ria
ble
s a
nd
Pa
ram
eter
s
15
TABLE 8
Table 8 shows the impact of changes in the baseline growth rate. It takes Symetra Financial’s growth rate
of 8% and reduces it by half to 4%. Compared to Scenario 5, the difference in NI falls by 1% in year one
and is equivalent in year five, the differences in ROA and ROE actually increase by .1% and 1.2% in year
one and by .1% and .4% in year five respectively. Compared to the parameter changes in the previous
scenarios, the impact of reducing the growth rate by half is relatively small.
Scenario 8: Symetra FinancialReduced Baseline Growth Rate
Baseline Growth Rate (%) 4%
Pre-Disaster Assets (2009) 22,226
Pre-Disaster Equity (2009) 1,481
Operating Leverage (%) 5%
Projected Decline in Demand (%) 20%
Duration of Demand Decline (years) 3
Duration of Facility Closing (days) 60
Facility Size (square feet) 5,000
Decontamination Costs (per square foot) 5.70
Tax Rate (%) 15%
Income Statements ($ in Thousands) 2008 2009 2010E 2011E 2012E 2013E 2014E
Sales (Net) 1,451.10 2,245.63 1,561 1,891 1,966 2,454 2,552
Cost of Goods Sold 1,114.60 1,568.69 1,145 1,386 1,442 1,799 1,871
Gross Profit 336.50 676.94 416 504 524 654 681
Selling, General, & Admin Expenses 265.80 328.59 261 312 324 400 416
Depreciation, Depletion, & Amortization 25.80 63.70 64 64 64 64 64
Operating Income 44.90 284.65 92 128 136 190 201
Interest Expense 31.90 31.90 32 32 32 32 32
Minority Interest in Income 0.00 0.00 0 0 0 0 0
Non-Operating Income (Expenses) 0.00 0.00 0 0 0 0 0
Pretax Income 13.00 252.75 60 96 105 158 169
Income Taxes 9.10 68.95 9 14 16 24 25
Income Before Extraordinary Items & Discontinued
Operations 22.10 183.80 51 82 89 135 144
Extraordinary Items and Discontinued Operations (net of
income taxes) 0.00 0.00 29 0 0 0 0
Post-Disaster Net Income 22.10 183.80 22 82 89 135 144
Baseline Net Income 22.10 183.80 204 216 229 242 255
Difference -89% -62% -61% -44% -44%
Post-Disaster Return on Assets (ROA) 0.8% 0.1% 0.4% 0.4% 0.6% 0.6%
Baseline Return on Assets (ROA) 0.8% 0.9% 1.0% 1.0% 1.1% 1.1%
Difference (baseline projection vs. post-disaster) 0.0% -0.8% -0.6% -0.6% -0.5% -0.5%
Post-Disaster Return on Equity (ROE) 12.4% 1.5% 5.2% 5.4% 7.5% 7.4%
Baseline Return on Equity (ROE) 12.4% 13.8% 12.8% 12.0% 11.3% 10.7%
Difference (baseline projection vs. post-disaster) 0.0% -12.3% -7.5% -6.6% -3.8% -3.3%
Implied Beta 1.19 -0.26 0.23 0.25 0.54 0.53
Va
ria
ble
s a
nd
Pa
ram
eter
s
16
APPENDICES. BACKGROUND CALCULATIONS
Appendix 1: Esterline Technologies
Disaster 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 317,582.16 359,675.52 424,468.94 499,905.00 575,341.06
Total Assets 517,547.00 552,570.16 594,663.52 659,456.94 734,893.00 810,329.06
Baseline 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 340,474.59 398,390.19 466,142.66 544,912.43 636,021.59
Total Assets 517,547.00 575,462.59 643,215.06 721,984.84 813,094.00 904,203.16
Baseline Income Statement
2009 2010 2011E 2012E 2013E 2014E 2015E
281,492 305,320 341,958 382,993 428,953 480,427 538,078
190,832 202,078 226,327 253,487 283,905 317,974 356,130
90,660 103,242 115,631 129,507 145,048 162,453 181,948
33,264 37,730 37,070 40,613 44,581 49,025 54,002
13,833 13,928 13,928 13,928 13,928 13,928 13,928
43,563 51,584 64,633 74,966 86,539 99,500 114,017
5,738 6,636 6,636 6,636 6,636 6,636 6,636
327 192 192 192 192 192 192
-1,594 2 2 2 2 2 2
35,904 44,758 57,803 68,136 79,709 92,670 107,187
2,510 4,901 8,670 10,220 11,956 13,901 16,078
33,394 39,857 49,133 57,916 67,752 78,770 91,109
0 0 0 0 0 0 0
- - - - - - -
33,394 39,857 49,133 57,916 67,752 78,770 91,109
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 26,611 30,184 COGS 190,832 202,078
% of Revenue 0.0945 0.0989 % of Revenue 0.6779 0.6619
Weighted Avg 0.0967 Weighted Avg 0.6699
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 327 192
% of Revenue 0.0012 0.0006
Weighted Avg 0.0009
17
Appendix 2: Esterline Technologies
Disaster 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 271,775.93 263,072.33 256,697.74 252,931.63 249,165.53
Total Assets 517,547.00 506,763.93 498,060.33 491,685.74 487,919.63 484,153.53
Baseline 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 340,474.59 398,390.19 466,142.66 544,912.43 636,021.59
Total Assets 517,547.00 575,462.59 643,215.06 721,984.84 813,094.00 904,203.16
Baseline Income Statement
2009 2010 2011E 2012E 2013E 2014E 2015E
281,492 305,320 341,958 382,993 428,953 480,427 538,078
190,832 202,078 226,327 253,487 283,905 317,974 356,130
90,660 103,242 115,631 129,507 145,048 162,453 181,948
33,264 37,730 37,070 40,613 44,581 49,025 54,002
13,833 13,928 13,928 13,928 13,928 13,928 13,928
43,563 51,584 64,633 74,966 86,539 99,500 114,017
5,738 6,636 6,636 6,636 6,636 6,636 6,636
327 192 192 192 192 192 192
-1,594 2 2 2 2 2 2
35,904 44,758 57,803 68,136 79,709 92,670 107,187
2,510 4,901 8,670 10,220 11,956 13,901 16,078
33,394 39,857 49,133 57,916 67,752 78,770 91,109
0 0 0 0 0 0 0
- - - - - - -
33,394 39,857 49,133 57,916 67,752 78,770 91,109
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 26,611 30,184 COGS 190,832 202,078
% of Revenue 0.0945 0.0989 % of Revenue 0.6779 0.6619
Weighted Avg 0.0967 Weighted Avg 0.6699
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 327 192
% of Revenue 0.0012 0.0006
Weighted Avg 0.0009
18
Appendix 3: Esterline Technologies
Disaster 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 316,717.21 360,150.91 433,365.11 520,541.50 607,717.90
Total Assets 517,547.00 551,705.21 595,138.91 668,353.11 755,529.50 842,705.90
Baseline 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 342,312.36 402,065.72 474,185.57 560,155.88 661,638.72
Total Assets 517,547.00 577,300.36 649,420.21 735,390.52 836,873.36 938,356.19
Baseline Income Statement
2009 2010 2011E 2012E 2013E 2014E 2015E
281,492 305,320 341,958 382,993 428,953 480,427 538,078
190,832 202,078 226,327 253,487 283,905 317,974 356,130
90,660 103,242 115,631 129,507 145,048 162,453 181,948
33,264 37,730 37,565 38,451 39,443 40,554 41,798
13,833 13,928 13,928 13,928 13,928 13,928 13,928
43,563 51,584 64,138 77,128 91,677 107,972 126,222
5,738 6,636 6,636 6,636 6,636 6,636 6,636
327 192 192 192 192 192 192
-1,594 2 2 2 2 2 2
35,904 44,758 57,308 70,298 84,847 101,142 119,392
2,510 4,901 8,596 10,545 12,727 15,171 17,909
33,394 39,857 48,712 59,753 72,120 85,970 101,483
0 0 0 0 0 0 0
- - - - - - -
33,394 39,857 48,712 59,753 72,120 85,970 101,483
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 6,653 7,546 COGS 190,832 202,078
% of Revenue 0.0236 0.0247 % of Revenue 0.6779 0.6619
Weighted Avg 0.0242 Weighted Avg 0.6699
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 327 192
% of Revenue 0.0012 0.0006
Weighted Avg 0.0009
19
Appendix 4: Esterline Technologies
Disaster 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 312,673.24 349,268.61 406,672.75 473,832.80 540,992.85
Total Assets 517,547.00 547,661.24 584,256.61 641,660.75 708,820.80 775,980.85
Baseline 2010 2011 2012 2013 2014 2015
Total Equity 282,559.00 340,474.59 398,390.19 466,142.66 544,912.43 636,021.59
Total Assets 517,547.00 575,462.59 643,215.06 721,984.84 813,094.00 904,203.16
Baseline Income Statement
2009 2010 2011E 2012E 2013E 2014E 2015E
281,492 305,320 341,958 382,993 428,953 480,427 538,078
190,832 202,078 226,327 253,487 283,905 317,974 356,130
90,660 103,242 115,631 129,507 145,048 162,453 181,948
33,264 37,730 37,070 40,613 44,581 49,025 54,002
13,833 13,928 13,928 13,928 13,928 13,928 13,928
43,563 51,584 64,633 74,966 86,539 99,500 114,017
5,738 6,636 6,636 6,636 6,636 6,636 6,636
327 192 192 192 192 192 192
-1,594 2 2 2 2 2 2
35,904 44,758 57,803 68,136 79,709 92,670 107,187
2,510 4,901 8,670 10,220 11,956 13,901 16,078
33,394 39,857 49,133 57,916 67,752 78,770 91,109
0 0 0 0 0 0 0
- - - - - - -
33,394 39,857 49,133 57,916 67,752 78,770 91,109
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 26,611 30,184 COGS 190,832 202,078
% of Revenue 0.0945 0.0989 % of Revenue 0.6779 0.6619
Weighted Avg 0.0967 Weighted Avg 0.6699
Duration of Facility Closing Factor 0.3288 Baseline 2009 2010
Minority Interest 327 192
% of Revenue 0.0012 0.0006
Weighted Avg 0.0009
20
Appendix 5: Symetra Financial
Disaster 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,576.18 1,687.14 1,859.12 2,052.48 2,245.84
Total Assets 22,226.00 22,321.68 22,432.64 22,604.62 22,797.98 22,991.34
Baseline 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,733.78 1,987.05 2,268.21 2,579.47 2,923.26
Total Assets 22,226.00 22,479.28 22,760.43 23,071.70 23,415.48 23,759.27
Baseline Income Statement
2008 2009 2010E 2011E 2012E 2013E 2014E
1,451 2,246 2,425 2,619 2,829 3,055 3,300
1,115 1,569 1,694 1,830 1,976 2,134 2,305
337 677 731 790 853 921 995
266 329 368 396 426 459 495
26 64 64 64 64 64 64
45 285 300 330 363 398 436
32 32 32 32 32 32 32
0 0 0 0 0 0 0
0 0 0 0 0 0 0
13 253 268 298 331 366 404
-9 69 40 45 50 55 61
22 184 227 253 281 311 344
0 0 0 0 0 0 0
- - - - - - -
22 184 227 253 281 311 344
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 253 312 COGS 1,115 1,569
% of Revenue 0.1740 0.1390 % of Revenue 0.7681 0.6986
Weighted Avg 0.1565 Weighted Avg 0.7333
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 0 0
% of Revenue 0.0000 0.0000
Weighted Avg 0.0000
21
Appendix 6: Symetra Financial
Disaster 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,424.62 1,373.23 1,326.69 1,285.40 1,244.10
Total Assets 22,226.00 22,170.12 22,118.73 22,072.19 22,030.90 21,989.60
Baseline 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,733.78 1,987.05 2,268.21 2,579.47 2,923.26
Total Assets 22,226.00 22,479.28 22,760.43 23,071.70 23,415.48 23,759.27
Baseline Income Statement
2008 2009 2010E 2011E 2012E 2013E 2014E
1,451 2,246 2,425 2,619 2,829 3,055 3,300
1,115 1,569 1,694 1,830 1,976 2,134 2,305
337 677 731 790 853 921 995
266 329 368 396 426 459 495
26 64 64 64 64 64 64
45 285 300 330 363 398 436
32 32 32 32 32 32 32
0 0 0 0 0 0 0
0 0 0 0 0 0 0
13 253 268 298 331 366 404
-9 69 40 45 50 55 61
22 184 227 253 281 311 344
0 0 0 0 0 0 0
- - - - - - -
22 184 227 253 281 311 344
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 253 312 COGS 1,115 1,569
% of Revenue 0.1740 0.1390 % of Revenue 0.7681 0.6986
Weighted Avg 0.1565 Weighted Avg 0.7333
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 0 0
% of Revenue 0.0000 0.0000
Weighted Avg 0.0000
22
Appendix 7: Symetra Financial
Disaster 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,576.18 1,687.14 1,859.12 2,052.48 2,245.84
Total Assets 22,226.00 22,321.68 22,432.64 22,604.62 22,797.98 22,991.34
Baseline 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,733.78 1,987.05 2,268.21 2,579.47 2,923.26
Total Assets 22,226.00 22,479.28 22,760.43 23,071.70 23,415.48 23,759.27
Baseline Income Statement
2008 2009 2010E 2011E 2012E 2013E 2014E
1,451 2,246 2,425 2,619 2,829 3,055 3,300
1,115 1,569 1,694 1,830 1,976 2,134 2,305
337 677 731 790 853 921 995
266 329 368 396 426 459 495
26 64 64 64 64 64 64
45 285 300 330 363 398 436
32 32 32 32 32 32 32
0 0 0 0 0 0 0
0 0 0 0 0 0 0
13 253 268 298 331 366 404
-9 69 40 45 50 55 61
22 184 227 253 281 311 344
0 0 0 0 0 0 0
- - - - - - -
22 184 227 253 281 311 344
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 253 312 COGS 1,115 1,569
% of Revenue 0.1740 0.1390 % of Revenue 0.7681 0.6986
Weighted Avg 0.1565 Weighted Avg 0.7333
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 0 0
% of Revenue 0.0000 0.0000
Weighted Avg 0.0000
23
Appendix 8: Symetra Financial
Disaster 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,562.30 1,651.19 1,785.73 1,929.46 2,073.19
Total Assets 22,226.00 22,307.80 22,396.69 22,531.23 22,674.96 22,818.69
Baseline 2009 2010 2011 2012 2013 2014
Total Equity 1,480.50 1,696.93 1,913.36 2,142.26 2,384.13 2,639.47
Total Assets 22,226.00 22,442.43 22,671.33 22,913.19 23,168.54 23,423.89
Baseline Income Statement
2008 2009 2010E 2011E 2012E 2013E 2014E
1,451 2,246 2,335 2,429 2,526 2,627 2,732
1,115 1,569 1,631 1,697 1,765 1,835 1,909
337 677 704 732 761 792 824
266 329 368 382 397 412 428
26 64 64 64 64 64 64
45 285 272 287 301 316 332
32 32 32 32 32 32 32
0 0 0 0 0 0 0
0 0 0 0 0 0 0
13 253 241 255 269 285 300
-9 69 36 38 40 43 45
22 184 204 216 229 242 255
0 0 0 0 0 0 0
- - - - - - -
22 184 204 216 229 242 255
Baseline 2009 2010 Baseline 2009 2010
Variable Costs 253 312 COGS 1,115 1,569
% of Revenue 0.1740 0.1390 % of Revenue 0.7681 0.6986
Weighted Avg 0.1565 Weighted Avg 0.7333
Duration of Facility Closing Factor 0.1644 Baseline 2009 2010
Minority Interest 0 0
% of Revenue 0.0000 0.0000
Weighted Avg 0.0000
24
REFERENCES
Balent, H., Carroll, G., Crothers, N., Johnson, R., and Lawes, S. 2003. Capitol Hill Anthrax Incident (GAO-03-686),
United States General Accounting Office.
Canter, D., Gunning, D., Rodgers, P., O'Connor, L., Traunero, C., and Kempter, C. 2009. “Remediation of Bacillus
Anthracis Contamination in the U.S. Department of Justice Mail Facility,” Biosecurity and Bioterrorism:
Biodefense Strategy, Practice, and Science 3(2): 119-127
Franco, C. & Bouri, N. 2010. Environmental decontamination following a large-scale bioterrorism attack: federal
progress and remaining gaps. Biosecurity and Bioterrorism: Biodefense Strategy, Practice, and Science8(2).
NRC (National Research Council). 2009. Reopening public facilities after a biological attack: A decision-making
framework. Committee on Standards and Policies for Decontaminating Public Facilities Affected by Exposure to
Harmful Biological Agents: How Clean is Safe? Washington, DC: National Academies Press.
Price, P. Anthrax sampling and decontamination: technology trade-offs. 2009. Lawrence Berkeley National
Laboratory. Retrieved from <http://escholarship.org/uc/item/1mm135gt>.
Revsine, L, D. W. Collins, W. B. Johnson, and H. F. Mittelstaedt. 2009. Financial Reporting and Analysis. Boston:
McGraw-Hill Irwin.
Rosenbaum, J., and J. Pearl. 2009. Investment Banking: Valuation, Leveraged Buyouts, and Mergers & Acquisitions.
New Jersey: John Wiley & Sons, Inc.
25
CHAPTER 5. CONCLUSION
by
Adam Rose
This report offers two major contributions to the analysis of economic impacts of an anthrax
attack on a major city in the United States. These are the impacts on residential real estate prices and the
impacts on the rate of return on business investment. Both factors affect decisions of residents and
businesses to leave a major city in the aftermath of such an attack.
A drop in real estate prices will lead to many residents owing more on their mortgages than they
have in equity. This status of an "upside-down" or "mortgage underwater" is a major source of voluntary
defaults or forced foreclosures. A significant reduction in the rate of return on investment is likely to
cause businesses to close, or at least shift their location or shift their activity to branches elsewhere
temporarily.
We develop two very distinct tools to be able to perform these analyses. A statistical model was
developed to predict the effects of anthrax attack on real estate prices. The model was based on cross-
sectional and time series data by individuals that code in the Seattle area, and the results yielded a very
good statistical fit. The model was applied to changes in variables representing general economic and
housing market conditions to ascertain the impacts of an anthrax attack on residential property values.
The property value changes were then inserted into a mortgage default model to estimate the extent to
which mortgages were underwater and the likelihood of default or foreclosure. The results for our worst
case scenario indicate that residential property values could decrease by over $50 billion, or a 33 percent
overall drop. Moreover, this increases the amount of negative equity by more than $15 billion. This
could result in more than 70,000 residential units being foreclosed.
An income statement simulation spreadsheet was developed to analyze the effects on business
rates of return. The program includes standard financial variables and arranges them in such a way that
the user can input variations affected by the anthrax attack. The major variables are the level and duration
of the projected region-wide economic downturn. Our simulation results for the worst case scenario
indicate that the average business may face several years of negative returns on equity and assets
following the anthrax attack. Businesses that are the most vulnerable are those confined to the Seattle
area with no branch plants or parent company connections that help can cushion the financial shock.
26
The two sets of results, combined with the analysis of the companion study by John et al. (2011),
can be used by policy-makers to ascertain the level of subsidies and other financial incentives to induce
population to stay or return to a major city affected by an anthrax attack or other major type of terrorist
attack that would cause significant damage and require extensive decontamination.
We emphasize the caveat that our results are intended to be illustrative rather than precise figures
for policy formulation at this time. While the predictive models are statistically sound, their application
could still be improved. For the real estate model, this would entail obtaining additional data on the
distribution of original mortgage levels and pre-attack equity levels in each of the three geographic zones
in Seattle. The analysis could be further bolstered by finding ways to differentiate the prediction of
median home values by impact zone. The final step would be to improve the supplementary analysis of
the likelihood of default/foreclosure in the event that home values drop significantly below mortgage
balance levels. Unfortunately, there are few real world examples upon which to base this analysis. For
the business valuation model, improvements could be made by the inclusion of more firm-specific data on
operating leverage. In addition, the model is inherently linear, and thus currently does not capture likely
non-linearities of reactions to extreme events.
1