Project Three:
Joining the
Conversation
Free Write
Should smoking be
allowed anywhere
on campus?
Project 3:
Support An
Arguable Claim
1000 -1200 words
Expectations
For
Project 3
Final
THE BIG PAPER
PEER REVIEW
Background/ context on topic, evidence,
counterarguments, and conclusion that
offers the reader something for further
thought.
You will do this next week
Complemented by a presentation.
Due last week of Class
PRESENTATION
4-6 minutes long, presented in small
groups
FINAL DRAFT
1,000 - 1,200 words, and it will include a
thesis and use in-text citations -- Must
include AT LEAST your six sources
Early Draft
Due this Sunday, Nov.
12
Early Draft
The early draft should
be a 700 word draft that
argues in support of a
particular thesis, uses
at least three sources,
and includes both an
introduction and
conclusion. Instructors
will provide extensive
feedback on this Early
Draft.
Introduction will do these things:
1. Introduce the topic
2. State why the topic is important
3. State that there is a difference of opinion about this topic
4. Clearly articulated thesis
Should include:
1.Summary of works being discussed
2.Definition of key terms
3.Explanation of key theories
“
A good argumentative thesis will do the following:
Assertive language
Demonstrate authority on the subject
Address the opposing view
Act as a “roadmap” for the argument by listing
supporting points
BODY GRAPHS:
Topic Sentence: The topic sentence should identify the main idea and point of
the paragraph. To choose an appropriate topic sentence, read the paragraph and
think about its main idea and point.
Context: The context (1-2 sentences) should mention the author(s) and
his/her/their article or book title.
Quote: aim for 1 - 2 lines
Analysis: The analysis should accomplish 3 tasks:
1.Summarize the quote and article author’s point of view
2.Relate the quote back to the overall conversation
3.Support the student’s opinion about the topic
Concluding Sentence: 1 sentence suggesting how the source contributes to the
student’s argument
The P3 essay must include a fair examination of the
counterargument.
How do I include
counterargument?
What do we mean by “fair”?
The other side must be examined without unfair
judgement (ie, “All religious
people are irrational”)
At least one credible source must be used to justify the
counterargument’s
viewpoint
The author must be able to logically point to
weaknesses in the
counterargument without using logical fallacies
Pitfalls:
Avoid fallacies
Effective presentation design (this presentation is an
EXAMPLE OF WHAT NOT TO DO
Homework
⊙ Homework: Write Intro Graph and
Prepare to Read Aloud to Class
⊙ Bring Laptop
Wednesday (In Class):
Workshop, Fallacy, Conference Sign-Up,
Presentation
Hurricane Damage to the Ocean Economy
Hurricane Damage to the Ocean Economy
Hurricane damage to the ocean
economy in the U.S. gulf region in 2005
Counties and parishes of the gulf coast ocean economy affected
by Hurricanes Katrina and Rita saw the greatest insured dollar losses
in 1 year from suchlike catastrophes in U.S. history; the region
has yet to recover fully a year after the maelstrom
Charles S. Colgan
and
Jefferey Adkins
Charles S. Colgan is a
professor in the
National Ocean
Economics Program,
Muskie School of
Public Service,
University of Southern
Maine, Portland,
Maine, and chief
economist for Market
Data for the National
Ocean Edonomics
Program; Jefferey
Adkins is an
economist and
program manager
with the National
Oceanic and
Atmospheric
Administration’s
Coastal Services,
Charleston, SC.
76
I
n 2005, insured losses from hurricanes and
other catastrophes were greater than in any
other year in U.S. history. NOAA’s National
Hurricane Center estimates that $85 billion of
total damages resulted from Hurricanes Katrina
and Rita alone. One year later, the region affected
by these two hurricanes still struggles to recover,
both as a place to live and as a viable economy.
Using data from the BLS Quarterly Census of
Employment and Wages, the National Ocean
Economics Program has developed a data series
that allows the economic damage to coastal
regions to be seen in a new light: what happens
to the economic value derived from the ocean
when the ocean turns from resource and respite
to a massive engine of destruction?1
According to Federal disaster declarations,
Hurricane Katrina affected the entire States of
Mississippi and Louisiana, plus 22 counties in
Alabama and 9 in Florida. Rita affected all of
Louisiana, plus 26 counties in Texas. The greatest
effects were in the counties (parishes in
Louisiana) closest to the coast, where the storm’s
effects were at their maximum intensity. Coastal
counties and parishes include those designated
as such by each State under the Federal Coastal
Zone Management Program, as well as those
designated as coastal watershed counties or
parishes by the U.S. Geological Survey.
Virtually all of the coastal zone and watershed
counties2 or parishes of Alabama, Mississippi,
and Louisiana, plus the coastal counties in Texas
from Houston eastward, were affected by the two
hurricanes. The coastal zone counties or
parishes of the four States account for nearly a
Monthly Labor Review
Augustl 2006
quarter of employment and wages in those
States. In Louisiana, the coastal parishes are
more than half of the State’s economy. The
combined coastal zone and watershed counties
and parishes on the Gulf of Mexico constituted
14 percent of employment in Alabama, 4 percent
in Mississippi, 6 percent in Florida, a considerably greater 33 percent in Texas, and fully
80 percent in Louisiana.
The ocean economy is defined as industries in
marine construction, living resources (seafood
processing and marketing, plus aquaculture),
shipbuilding and boatbuilding, minerals (primarily oil and gas exploration and production),
marine transportation and related goods and
services, and, finally, tourism and recreation
industries whose establishments are located close
to the shore of the ocean or the Great Lakes.
In 2004, the ocean economy of the region
encompassing Florida, Alabama, Mississippi,
Louisiana, and Texas, stretching from Franklin
County, Florida, to Brazoria County, Texas, employed 291,830 people in wage and salary jobs
paying nearly $7.7 billion in wages. (See table
1.) The affected States accounted for 13 percent
of employment and wages in the U.S. ocean
economy.
However, these gross figures mask a key fact
about the region: it is the industrial heartland of
the U.S. ocean economy. As the following
tabulation shows, the region accounts for more
than a third of U.S. employment in marine
construction, more than a fifth of employment in
fisheries (living resources), shipbuilding, and
boatbuilding, and more than half of employment
Table 1. Employment and wages in ocean economy industries in the Gulf of Mexico region, 2004
2004
Sector and industry
Employment
2004
Wages
(millions of
dollars)
Ocean economy, total ...............
Construction ...........................................
Living resources ....................................
Fish hatcheries and aquaculture ........
Fishing ................................................
Seafood markets ................................
Seafood processing ............................
Minerals ..................................................
Shipbuilding and boatbuilding ................
Boatbuilding and repair ......................
Shipbuilding and repair .......................
291,830
12,094
12,552
1,653
571
1,686
8,642
15,105
l35,839
3,567
32,272
$7,694.8
548.1
251.9
37.7
10.8
32.0
171.4
1,077.0
1,443.2
125.5
1,317.7
Tourism and recreation ............................
Amusement and recreation services ..
Boat dealers ........................................
178,404
4,150
2,078
2,716.7
74.9
73.5
Chart 1.
Sector and industry
Employment
Wages
(millions of
dollars)
Eating and drinking places .....................
Hotels and lodging places ......................
Marinas ...................................................
Recreational vehicles in parks
and campsites ......................................
Scenic water tours ..................................
Sporting goods ........................................
Zoos and aquariums ..............................
131,985
34,870
1,447
1,718.6
725.5
38.2
470
1,676
106
1,622
8.9
43.0
2.9
31.1
Transportation .........................................
Deep-sea freight transportation ..........
Marine passenger transportation ........
Marine transportation services ...........
Search and navigation equipment ......
Warehousing ........................................
37,836
3,109
375
28,485
2,251
3,616
1,657.9
226.0
20.6
1,163.8
107.1
140.4
Loss of employment in the ocean economy industries as a result of Hurricanes Katrina
and Rita in 2005
Percent of
employment lost
10–27
5–9
3–4
2
0–1
Monthly Labor Review
August
2006
77
Hurricane Damage to the Ocean Economy
in the ocean-related component of oil and gas exploration
and production:
Percent of U.S. ocean economy
Employment
Construction .............................
Living resources ........................
Offshore oil and gas .................
Shipbuilding and boatbuilding ..
Tourism and recreation .............
Transportation .........................
38.8
20.2
51.1
22.0
10.6
13.3
Wages
34.8
13.7
49.3
19.1
8.9
9.1
The region also accounts for a disproportionate share of
marine transportation-related employment.
Chart 1 shows the counties and parishes bordering the
Gulf of Mexico that were declared disaster areas as a result of
Hurricanes Katrina and Rita in 2005. The shading represents
the portion of employment in each county that was accounted
for by construction, living resources, minerals, shipbuilding
and boatbuilding, and transportation. The heaviest concentration of these industries extends from Jackson County,
Mississippi, to Cameron Parish, Louisiana. In these counties,
the portions of the ocean industry other than tourism and
recreation range from 3 percent to 27 percent of county
employment.
The economic effects of these hurricanes have focused
on discussions of the losses in New Orleans, perhaps the
largest disaster effects on a major American city since the
San Francisco earthquake a century ago. But the economy
affected was significant to the nation not only because of the
loss of the unique charms of the Crescent City. The affected
region was the heart of the industrial sectors of the American
ocean economy, and the recovery of these industries will be
critical to both the region and the Nation.
Note
1
For information on the definitions of the ocean economy, visit
www.oceaneconomics.org, the Web site of the National Ocean Economics
Program.
2
Coastal zone counties are those within a State’s defined coastal
zone management program. Watershed counties are defined by the
U.S. Geological Survey.
LABSTAT available via World Wide Web
LABSTAT, the Bureau of Labor Statistics public database, provides current and historical data for many BLS surveys, as well as numerous news releases.
Data can be accessed by using the data retrieval tools available at
http://www.bls.gov/data
If you have questions or comments regarding the LABSTAT system on the Internet,
address e-mail to
labstat.helpdesk@bls.gov
78
Monthly Labor Review
Augustl 2006
Ann Reg Sci (2014) 52:325–342
DOI 10.1007/s00168-013-0587-8
ORIGINAL PAPER
When oceans attack: assessing the impact of hurricanes
on localized taxable sales
Ariel R. Belasen · Chifeng Dai
Received: 16 April 2012 / Accepted: 6 December 2013 / Published online: 21 December 2013
© Springer-Verlag Berlin Heidelberg 2013
Abstract We examine the impact of hurricanes in Florida on county-level taxable
sales revenues. Conditional on the strength of the hurricane, within 6 months after a
hurricane strikes a county, revenues decline as much as 17 %, whereas revenues in
neighboring counties increase by upward of 17 % over that same time frame. This
decline in revenue is found to be dependent on the commercial makeup of a hurricanestricken county. Particular focus is given to tourism-related subsectors within the local
economy. Finally, we show that along the pathways of hurricanes, initially hit counties
face a more severe burden, ranging as high as a 33 % immediate decline in taxable
revenues in 1 month for coastal counties. As the hurricane weakens, the direct impact
is lessened; however, there is evidence of spillover damage in neighboring areas.
JEL Classification
H71 · R11
1 Introduction
Hurricanes regularly make landfall in the USA. The destructive power of hurricanes
can wipe out thousands of lives and cause billions of dollars worth of infrastructure
and private property losses, as seen in the aftermath of Hurricane Katrina and more
recently Hurricane Sandy. The forced migration prompted by hurricanes also creates
A. R. Belasen (B)
Department of Economics and Finance, Southern Illinois University Edwardsville,
Edwardsville, IL 62026, USA
e-mail: abelase@siue.edu
C. Dai
Department of Economics, Southern Illinois University Carbondale,
Carbondale, IL 62901, USA
e-mail: daic@siu.edu
123
326
A. R. Belasen, C. Dai
the potential for an area to lose its human capital. This loss of physical and human
capital in a region can have significant short-term and possibly long-term effects on
regional economic growth. In addition to their damage to physical and human capital,
hurricanes also disrupt regional economic activity due to temporary lack of utility service, employee absenteeism, supply chain interruption, etc. Understanding the effects
of hurricanes on a regional economy is critical to mitigate and recover the cost caused
by hurricane damage. Policymakers and local businesses could make projections of the
potential impacts before actual disasters occur for mitigation strategies and develop
better recovery plans (Ewing et al. 2009).
Most existing studies of hurricanes and other natural disasters have been conducted
in the framework of a singular event study. For example, Guimaraes et al. (1993) study
the deviation that Hurricane Hugo caused South Carolina to take from its general
trend of economic growth using a with-and-without comparison of the economy. West
and Lenze (1994) present a general framework to estimate the regional economic
impact of natural disasters using primary and secondary data and apply their research
to the problem of estimating the impact of Hurricane Andrew on the economy of
Florida. Using time-series econometric models, Ewing and Kruse (2001) study the
effect of Hurricane Bertha on unemployment in Wilmington, NC; Ewing et al. (2005)
analyze the effect of Hurricane Bret on unemployment in Corpus Christi, Texas, as
well as a subsequent study on Tornadoes in Oklahoma City (Ewing et al. 2009). Xiao
(2011) and Xiao and Feser (2013) use time-series analysis under a quasi-experimental
pairwise matching design to examine local economic impacts of the 1993 Midwest
flood. Ewing and Kruse (2002) were among the first to examine a series of hurricanes
within a singular framework, namely five hurricanes that struck Wilmington, NC,
during the late 1990s. They made use of a time series analysis to look at the effects
of each of five hurricanes, both individually and collectively, on the unemployment
rate of Wilmington, North Carolina. However, studies of multiple disaster events over
large geographic areas are rare.
Rose et al. (1997) estimate both the direct and indirect regional economic losses
from earthquake-damaged electric utility lifelines using specially designed input–
output and linear programming models. Extending the previous study, Rose and Liao
(2005) examine how damages from disasters are muted by the inherent and adaptive resilience of individual businesses and of regional markets. They advance the
computable general equilibrium (CGE) analysis to reflect resilience and to provide
a more accurate estimate of impacts by disasters. On the other hand, using survey
methodology, Kapucu (2008) and Kim and Kang (2010) show that it can be difficult
at times for governments to convey the seriousness of the impending disaster and that
household preparedness depends on the individual household’s access to communication resources. Furthermore, Cho et al. (2001) integrate transportation network and
regional economic models to estimate the economic loss due to earthquake impacts
on transportation and industrial capacity.
These studies provide important insights on how a particular hurricane or a series
of hurricanes affect a particular local economy. However, in general, the effects of
hurricanes vary depending on the severity and the timing of the storm and the characteristics of the local economy. Consequently, the insights obtained through an analysis
of one hurricane or multiple hurricanes on one local economy may not be useful in
123
Assessing the impact of hurricanes on localized taxable sales
327
generalizing to a different hurricane or a different local economy. Moreover, some of
these studies fail to account for other macroeconomic changes that coincide with the
dates of their natural disasters and therefore could end up with biased estimates of the
actual impact of the disasters. In addition, many of these studies (e.g., Rose et al. 1997;
Rose and Liao 2005; and Cho et al. 2001) use hypothetical, stylized natural disasters
and therefore consider only limited aspects of the complex set of events that actually
occur during a natural disaster.
This research contributes to the literature by examining changes in localized taxable
sales revenues caused by the devastating effects of a series of hurricanes that actually
struck Florida between 1992 and 2006. We pay particular attention to specific timing
effects in terms of both the paths taken by these hurricanes and the duration of the
economic impacts brought on by them, while controlling for external macroeconomic
events to isolate the average impact of hurricanes on revenues. We also specifically
address local geography to assess the hurricanes as they weaken while moving inland.
The current study is closely related to Belasen and Polachek (2008, 2009) who
use the generalized difference in difference (GDD) technique to examine a series of
hurricanes that hit the state of Florida to identify potential exogenous shocks in the
labor market using quarterly data. Belasen and Polachek (2009) found that across a
two-year duration, wages tended to rise while employment tended to fall in hurricanestricken counties relative to unaffected counties.
Most hurricanes that strike the USA occur in the Gulf States and the southeastern
states. Florida is a member of both subsets of states, and more hurricanes hit Florida
than any other US state. In the 15-year period between 1992 and 2006, a total of 18
hurricanes struck Florida, and none of Florida’s 67 counties have escaped the effects
of these hurricanes. Five of the six most damaging Atlantic hurricanes of all time have
struck Florida over the course of this time period along with a number of relatively
weak hurricanes as well as providing a fairly comprehensive sampling of various
strength hurricanes.
Since each of these 18 hurricanes in our study varies in the exact timing, strength,
path, and the degree of damage and therefore can be analyzed independently or collectively, they fit into multiple random experimental groups (counties that have been
hit by a hurricane in a given month) as well as multiple control groups (counties that
have not been hit by a hurricane). By comparing those counties that have been hit to
the other counties that were not across multiple hurricanes and time periods, we are
able to investigate how the impact of a hurricane differs by both the characteristics of
the hurricane and the characteristics of the local economy. Moreover, the estimation
techniques address the concern of any macroeconomic changes that coincide with
the timelines of the hurricanes by accounting for statewide trends in the data. In this
way, we can better isolate the outcome of a hurricane as well as quantify its economic
impact over time.
Furthermore, hurricanes that strike a specific county directly can also have an impact
on the economies of neighboring counties. Thus, similar to Belasen and Polachek
(2008, 2009) we study the spillover effects of hurricanes on neighboring counties
as well. In contrast to the studies based on hypothetical natural disasters (e.g., Rose
et al. 1997; Rose and Liao 2005; and Cho et al. 2001), our analysis is based on the
total effect of actual hurricanes on county-level taxable sales. While our analysis does
123
328
A. R. Belasen, C. Dai
not explicitly address the effect of a specific hurricane-caused event (e.g., electricity
interruption, transportation network damage, or recovery activities), it captures the
total effect of the set of complex events that occur during and after hurricanes. Finally,
while we are unable to discern a pure hurricane effect separate from recovery efforts,
by identifying impacts at multiple subsequent time periods, we can identify the scope
of the recovery relative to the initial impact.
Our analysis focuses on the impact of hurricanes on county-level taxable sales revenue, ranging from the sales of groceries, automobiles, hotel rooms, utilities telephone
services, to banking services. Taxable sales are not only key indicators of economic
conditions but also important tax bases for state and local governments. Since most
states and local governments base their budgets on their revenue forecasts, strong and
viable tax revenue estimates are vital to regional economies recovering from disasters.
Understanding the effect of hurricanes on taxable sales helps local policymakers and
fiscal analysts make better revenue projection and better expenditure decisions to facilitate the recovery process. And for Florida, in particular, as a state without an income
tax, the stability of taxable sales revenue is of upmost importance for long-term budget
plans. In addition, knowledge about hurricane-related costs also helps local businesses
create better economic plans before and after hurricanes.
Hurricanes can have very different effects on different sectors of a local economy.
In view of the fact that tourism is a mainstay of Florida’s economy, it is also interesting
to study how hurricanes affect Florida’s tourism industries. Therefore, in addition to
the overall sales of all categories subject to Florida sales and use taxes, we examine
the taxable sales of two tourism-related industries, i.e., the taxable sales of restaurants
and the taxable sales of bars and taverns.1
The current study complements Belasen and Polachek (2008, 2009) by analyzing
the impact of the same group of hurricanes on taxable sales revenues across a similar
period of time. For states that do not collect state and local income tax, such as Florida,
sales and usage taxes are the main sources of revenues for state and local governments.
Therefore, taxable sales data are better indicators of state and local tax revenues for
such states than employment and wages.
Furthermore, we make use of monthly county-level sales data in our analysis,
which allows us to better distinguish short-term and long-term effects of hurricanes.
For example, we find that when a weaker hurricane directly hits a county, taxable sales
revenues relative to unaffected counties will immediately decline by more than 7 %
on average in that month; however, revenues may begin to rebound in the month that
follows. On the other hand, revenues will show no significant change if that quarter is
examined as a whole. Meanwhile, stronger hurricanes will have persistent destructive
effects that may lead to a 20 % or greater decline in taxable revenues over that same
time frame. Finally, in contrast to Belasen and Polachek (2009), by tracking the paths
of hurricanes, we find that the impact of hurricanes on local economies depends on
the physical locations of the local economies, particularly in relation to the entry and
exit points of the hurricanes.
1 Unfortunately, data limitations prevent us from being able to examine the tourism sector as a whole.
123
Assessing the impact of hurricanes on localized taxable sales
329
2 Econometric model
In estimating the impact of hurricanes on Florida’s labor market, Belasen and Polachek
(2008, 2009) utilized the GDD approach to examine changes in the growth rates of
employment and earnings in the average county hit by a hurricane relative to the
average county that was not hit. For the purpose of our analysis, it will be more
instructive to examine the changes in revenue levels relative to the average as opposed
to the changes in growth rates so that the interpretation of the estimated coefficients
will have a more practical purpose for policymakers. With that in mind, the model
we use in this study will comprise similar controls to the model used in Belasen and
Polachek (2009); hence, we will need to use an econometric approach which will
control for seasonality, physical and demographic characteristics of each county, as
well as the relative wealth of each county. To facilitate this, we will employ a fixed
effects (FE) model to estimate the following:
+ HitD δit + H N εit + u it
ln Rit = β0 + αi + β1 T + Mγ
it
(1)
Here, R captures the level of taxable revenue collected by private firms in Florida
county i in time t. We use the natural logarithm of R for ease of interpretation. αi
captures any time-invariant county-specific factors such as geographical location and
physical size. Furthermore, seasonal trends are captured by the vector M which is a
collection of individual monthly dummy variables to control for seasonality, and a
time control vector, T , is included to capture the overall time trend. Finally, the factors of interest are the vectors H D and H N , which examine the impact of hurricanes
on counties inside and outside the locus of destruction. The D identifies directly hit
counties (i.e., counties within the locus of destruction), while the N identifies counties that neighbor the directly hit counties (i.e., counties within the outer band of the
hurricane). Each of these hurricane vectors is comprised of a split classification of
weaker and stronger hurricanes, separated by the maximum wind speeds. Category
1–3 hurricanes that have a wind speed under 100 mph are classified as weaker hurricanes, and category 4–5 hurricanes with wind speeds in excess of 100 mph are classified
as stronger hurricanes. Furthermore, following the findings in Ewing and Kruse (2002)
and Belasen and Polachek (2009),2 we supplement these hurricane variables with lags
to examine the specific timing of the economic shock following the hurricane strike.
To estimate the model with fixed effects, we use the mean-differenced county-level
value for revenue regressed on the time and month trends as well as the hurricane
effects, such that Eq. (1) becomes:
ln Rit − ln R̄i = (β0 − β0 ) + (αi − αi ) + β1 t + Mγ
+ HitD δit + H N εit + (u it − ū i )
it
(2)
2 Prior studies on hurricanes primarily used quarterly data and found significant results one quarter after
the strike (i.e., between 4 and 6 months following the hurricane). With monthly data, we will be examining
each of the 6 months after the strike. Xiao (2011) shows that local economies are relatively resilient and
will bounce back shortly after a disaster strikes.
123
330
A. R. Belasen, C. Dai
and the estimation of Eq. (2) reduces to the following form:
γ̂ + HitD δ̂it + H N ε̂it + Δû it
Δ ln R̂it = β̂1 t + M
it
(3)
For counties on the path of hurricanes, the demand for food, water, and other
disaster supplies often increases ahead of hurricanes, as households and businesses
prepare themselves and their properties for the potential damage of hurricanes. In the
aftermath of hurricanes, as the hurricane causes damages to infrastructure and business
properties and interrupts utility service and supply chains, supply often decreases
immediately in directly hit counties. Consumers in directly hit counties often increase
their spending on disaster supplies and are forced to eat out at restaurants or stay
in hotels due to loss of utility service or damage to their properties. However, they
often are forced out to neighboring areas to find stores, restaurants, and hotels that are
unaffected by hurricanes.
Over time, as recovery efforts restore utility service and replace damaged infrastructure and properties, we expect supply to steadily recover in affected counties. Moreover, as businesses and households start to repair the damages to their properties
financed by their saving accounts and relief funds, we expect the demand for building
supplies and construction services to increase significantly in directly hit counties. We
expect this increase in demand to have a positive spillover effect on neighboring areas
until the supply in directly hit counties completely recovers.
Therefore, our hypothesis is that revenues will decrease immediately in the month
when hurricanes strike a county. In addition, we expect to see the impact become muted
or even positive as we examine counties further out from the locus. Over time, we
expect to see revenue gradually recover for both directly hit counties and neighboring
counties. We also believe that the impact will vary across different business sectors for
the affected counties. Moreover, following Belasen and Polachek (2009), we anticipate
that increases in the relative strength of the hurricanes will exacerbate this impact.
Among the controls, the monthly dummies should capture the seasonality in revenues by showing somewhat of a downturn in summer months when tourism declines
and a jump in the winter months when it picks backup. This is particularly important to control for since the hurricane season coincides with the seasonal decrease in
tourism. Finally, the time trend should come out significantly and positively, reflecting
the overall growth in sales revenues over the sample period.
3 Data
The hurricane data used in this study come from hurricane reports from the National
Hurricane Center of the National Oceanic and Atmospheric Administration (NOAA).
NOAA is a federal agency within the Department of Commerce that examines the conditions of the oceans and the atmosphere. In particular, the NOAA evaluates ecosystems, climatic changes, weather and water cycles, and commerce and transportation.
NOAA reports that most hurricanes that strike the USA strike the Gulf States and
the southeastern states. In the 15-year period between 1992 and 2006, a total of 18
hurricanes struck Florida, and none of Florida’s 67 counties have escaped the effects
123
Assessing the impact of hurricanes on localized taxable sales
331
of these hurricanes.3 Details regarding these 18 hurricanes can be found in Table 1 in
the “Appendix.”
Hurricanes are classified by the Saffir–Simpson scale according to their maximum
wind speed, with weaker category 1 hurricanes having wind speeds between 74 and
95 mph and the strongest category 5 hurricanes having speeds greater than 155 mph.
Beyond that, the paths of the hurricanes were traced following NOAA coordinates,
and following Belasen and Polachek (2008), any county within a 50 km radius (or
100 km diameter) of the path of the hurricane was considered to be directly hit by the
hurricane.4 Those counties outside of that band but within a wider 500-km band were
considered to be counties neighboring the hurricane and would have faced a relatively
weaker impact of the hurricane than those within the locus of destruction.
The gross sales data are collected from the Gross and Taxable Sales Reports by
the Bureau of Economic and Business Research at the University of Florida (BEBR),
which are prepared based upon data provided by the Florida Department of Revenue.
The gross sales data include all sales reported monthly by businesses for sales and
use taxes as required by the state law. They range from the sales of retail, wholesale,
hotels, transportation, utility services, to insurance and banking services as classified
by the Florida Department of Revenue. The taxable sales data used in our analysis are
county-level data, spanning the time period starting from January 1992 and continuing
through December 2006. The average county’s taxable sales revenue is nearly $264
million; however, there is an extremely wide disparity between the different counties.
A summary of the sales data can be found in Table 2.5
4 Estimation results
Our findings reveal that after controlling for time and seasonality, hurricanes do in
fact decrease taxable revenues at several points in the 6 months following a hurricane
strike. Table 3 summarizes the results for the coefficients of interest from Eq. (3).
The results indicate that stronger hurricanes cause a much more pronounced and
enduring downturn in taxable revenues among directly hit counties. This difference
can also be seen when examining the neighboring counties as well. Weaker hurricanes actually have a positive net cumulative impact on sales revenues in neighboring
counties, presumably due to an increase in demand for goods and services as consumers in directly hit counties may be forced to shop outside their home county due
to hurricane-related closures. However, stronger hurricanes seem to cause a downturn
in neighboring counties that is similar in scope to the downturn caused by weaker hurricanes in directly hit counties. This indicates that even outside of the locus, stronger
3 Note that five of the six most damaging Atlantic Hurricanes of all time struck Florida during this time
period, as did some of the least damaging hurricanes. Therefore, this time period serves to provide a good
mix of the potential economic disruption of hurricanes.
4 Note that when expanding the radius to 100 or 150 km the results in our analysis did not have a statistically
significant change.
5 Note that in some months, the tax revenues were negative, reflecting months in which the State of Florida
received a net tax loss because of adjustments due to overpayments in previous time periods. While the
negative values are a relatively rare phenomenon, tax adjustments occur consistently throughout the data.
123
332
A. R. Belasen, C. Dai
hurricanes can still yield heavy destructive capabilities. This is consistent with the
findings in Hallegatte et al. (2008) that examined the 2004 and 2005 hurricanes used
in this study.
By aggregating the statistically significant results, we find that when a weaker
hurricane directly hits a county, taxable sales revenues relative to unaffected counties
will immediately decline by 7.18 % on average in that month; however, revenues
will increase by 0.40 % across the next 6 months following the strike. Relative to
the mean level of revenues, this implies an immediate decrease of $18.92 million
followed by a gain of $1.05 million across the next 6 months for an aggregate change
of −$17.87 million in revenues for the average county directly hit by a hurricane
relative to unaffected counties. Revenues in neighboring counties do not experience
an immediate change; however, they will eventually rise by 7.9 % on average relative
to unaffected counties. This reflects a revenue increase of $20.82 million.
While the changes in tax dollars collected only amount to a few million dollars at
the state level from weaker hurricanes, stronger hurricanes, on the other hand, have
a much more significant impact on tax revenues. Among directly hit counties, the
average impact is an immediate 9.81 % decline, which corresponds to a $25.85 million
decrease in taxable revenues. And in the following 6 months, the average aggregate
downturn increases by an additional 7.36 %. This reflects a total aggregate decrease
across the 7 months of $45.25 million in taxable revenues or a state-level decrease of
$2.7 million in tax dollars collected per directly affected county.6 Larger hurricanes
like Ivan which impact over a dozen counties throughout the duration of the hurricane
can therefore result in close to $33 million lost tax dollars for the state. Finally, as
stated earlier, the impact of stronger hurricanes on neighboring counties is similar to the
impact of weaker hurricanes on directly hit counties. There is no significant effect from
the hurricane immediately after impact, but the overall aggregate result is a 17.56 %
increase in revenues for the average neighboring county relative to unaffected counties.
Overall, we see evidence that indicates that a faster recovery is all the more important
for the stabilization of not only the local economy affected by the hurricane, but also the
state-level economy. Clukey (2010) points out that despite the best efforts of recovery
initiatives, there are drawbacks, especially to the individuals who are immersed in the
recovery efforts.
5 Expanding the analysis into subsectors of the local economy
To check for the specific severity of hurricane damage on tourism, we apply the
same model to economic subsectors that we believe will be particularly hard hit by a
hurricane, namely the restaurant sector and the bar & tavern sector.7 Currently, only
one in three Americans consumes one or more meals at home per day (Berry 2002).
6 While overall sales tax rates differ by locality in Florida, the state government collects a flat 6 % rate
across the entire state.
7 We do not include the hotel sector in our study of tourism-related subsectors. First, taxable sales by
the hotel sector are combined with those by apartments and rooming houses, etc., in the data provided by
the Florida Department of Revenue; second, the classification of this category was not consistent for the
duration of our analysis.
123
Assessing the impact of hurricanes on localized taxable sales
333
Therefore, we expect to see that major reductions in supply should heavily impact
taxable sales revenues among local restaurants and taverns even if the reductions are
only short-lived.8 The framework for the model remains the same except that now we
are looking specifically at those two subsectors.
Theoretically, unlike other sectors, local residents’ demand for food services could
increase immediately for directly hit counties as they are forced to eat out at restaurants
due to loss of utility service or damage to their properties; however, consumer demand
from tourists will decrease immediately for those counties due to a decrease in the
number of tourists. Meanwhile, supply could also be diminished for these counties.
Thus, the overall effect could depend on the severity of the damages. For neighboring
counties, consumer demand will likely increase, but supply will not be affected as
much. So we should expect a more positive effect relative to directly hit counties.
Once recovery efforts restore electricity and replace damaged infrastructure, both
demand and supply will start to pick up in directly hit counties, but not as much in
neighboring counties. Therefore, the results of these subsectors could be different from
of the overall economy, especially immediately after hurricanes. The results of this
specification are shown in Tables 4 and 5.9
The findings reveal that the subsectors do in fact differ quite a bit from the overall
retail sector. In fact, the results in Table 4 suggest that there are opposing impacts on
the supply and demand side such that the net impact of a hurricane is mitigated. In
the restaurant subsector, in particular, strong hurricanes do not significantly alter total
revenue at all, and weaker hurricanes create an immediate 8.5 % drop in revenues. This
means that the average impact of a weak hurricane is a $2 million decrease in countylevel taxable revenues relative to unaffected counties and a decrease of $120,000 in
the average taxes collected of per affected county. While this may appear surprising,
one must remember that strong hurricanes likely create a positive demand shock (as
people lose the ability to eat at home) and a negative supply shock (as restaurants
are forced to shut down due to damage and electricity loss) which may offset one
another.
Meanwhile, among neighboring counties, the net cumulative impact of weak hurricanes is actually a 13.8 % increase in revenues (leading to a $3.24 million increase)
over the 6 months following the hurricane, whereas stronger hurricanes appear to
have a net cumulative 91.5 % decrease in revenues (note that there is a lot of noise
in the results here because some of the counties had very low numbers reported for
restaurants—hence, one or two closures may have in fact constituted the majority of
the market in those rural counties).
8 Individuals in hurricane-stricken regions often report that the loss of electricity and other infrastructure
damage often forces them to actually increase restaurant consumption relative to meals at home, but that
they often are forced out to neighboring areas to find restaurants that were unaffected by the hurricane.
9 Note that the number of units per county diminishes greatly when examining the subsectors, and some
subsectors are completely unreported by the BEBR for certain counties due to the sample size being so
low that it would compromise privacy. Hence, the significance level is not nearly as high in these as in
the overall examination. So, while the stronger hurricanes came up insignificantly, we attribute that to the
smaller number of firms surveyed rather than to the actual impact of the hurricanes. Thus, the analysis in
this section will focus only on the weaker hurricanes.
123
334
A. R. Belasen, C. Dai
Next, looking specifically at the bar and tavern subsector, taxable revenues will
decrease by an average of $221,000 in counties directly hit by a weaker hurricane
(7.5 %), which leads to an average $13,000 decline in taxes collected in those counties.
Furthermore, we see no significant impact among stronger hurricanes, suggesting once
again that the supply and demand shocks are offsetting each other. Among neighboring
counties, the aggregate change in taxable revenues in bar and tavern subsector, there is
a net cumulative increase of 25.7 % in taxable revenues over the six months following
a weak hurricane, which means that tax dollars would increase by roughly $45,000
on average. Among stronger hurricanes, there is a negligible net cumulative −3.4 %
decrease, resulting in a $6,000 average decline in tax revenues across the 6-month
period.
6 Expanding the analysis into local geographical factors
Belasen and Polachek (2009) showed that within labor markets, geographical characteristics of counties were not useful predictors in determining the impact of hurricanes
on the labor market of a county. By tracking the paths of hurricanes across Florida
and then comparing coastal to inland counties, they found an average impact differential of 0.01 % on wages. A major limitation of this particular study was that Belasen
and Polachek (2009) did not examine the origination of the hurricanes relative to the
affected counties.
A hurricane will be strongest when it first hits land and then weaken as it moves
across the land such that its weakest point while it is in the State of Florida will be
at its exit point, which means a hurricane that strikes Florida from the Atlantic will
have very different effects on the Atlantic coastline than it will on the Gulf coastline.
Theoretically, one should expect to see that the entry point will be hit harder than
the exit point since the hurricane will have weakened over the few days of travel.
Furthermore, as hurricanes weaken, the storm cell tends to move slower, and therefore, while the direct damage will be lessened, the indirect impact of heavy rain will
likely be felt even more in neighboring counties. To address this disparity, we introduce a further breakdown of the hurricane data by examining entry and exit points.
Hurricanes are broken up into two groups: those that originate in the Atlantic Ocean
and those that originate in the Gulf of Mexico. Counties on the Atlantic coastline
are “entry counties” if they are hit with a hurricane originating the Atlantic and “exit
counties” if they are hit with a hurricane originating in the Gulf. Likewise, Gulf
counties are “entry” and “exit” based on the same criteria. Equation (4) contains this
breakdown:
+ β̂3 H D + β̂4 H N + β̂5 H
Δ ln R̂it = β̂1 t + β̂2 M
it
DEntry
N Entry
+ β̂6 Hit
+ β̂7 HitDExit + β̂8 HitN Exit + Δû it
(4)
The individual coefficients reflect marginal impacts of the average hurricane on a
specific geographic characteristic. This regression structure lets us examine the impact
of hurricanes on six distinct sets of counties relative to baseline counties in which all
of the hurricane coefficients take the value of zero. For example, when a hurricane
123
Assessing the impact of hurricanes on localized taxable sales
335
hits a county in the entry group, the coefficients of interest are the combination of
β3 and β5 . Likewise, inland counties that are hit next will take a value of zero for
the entry coefficient (β5 ). Finally, exit counties will face the combined impact of β3
and β7 . Neighboring counties can also be split into entry, inland, and exit groups.
As before, these results will be further split between weak and strong hurricanes to
better assess the average impact. The regression results for this analysis are shown in
Table 6.
If we only account for the statistically significant results, then among the weaker
hurricanes, there are no statistical differences between any of the three groups of
counties. Stronger hurricanes, on the other hand, have a much greater impact initially
but appear to dissipate as the storm moves away from the entry point. Entry counties
that are directly hit by the hurricane face an average of 33.42 % decline in taxable
revenues relative to unaffected counties. Inland counties face 18.80 % decline, and exit
counties face 9.54 % decline. Statistically speaking, the remaining groups of counties
do not appear to be any different from one another.10 As we stated previously, a
likely explanation for these swings in revenues stems from both demand- and supplyside market shocks as buyers are forced to leave devastated areas for nearby areas to
purchase goods.
7 Conclusion
Florida finds itself in a unique position among other southeastern US states in that
it faces a threat from hurricanes both from the Gulf of Mexico and from the North
Atlantic Ocean. This study identified the average impact of a hurricane on Florida’s
taxable sales revenue. Since the bulk of Florida’s state revenues come from sales
and usage taxes, a disruption of this income stream can be particularly damaging to
state-level services.
Our findings show that the impact of hurricanes on localized taxable sales depends
critically on the strength and path of hurricanes and the economic makeup of the
local economies. Furthermore, we find that the coastal counties that are hit by hurricanes as they first make landfall will be harder hit than the counties that are hit as
the hurricane exits the state. Additionally, by aggregating the statistically significant
results over a six-month period, we find that when a weaker hurricane directly hits a
county, taxable sales revenues relative to unaffected counties will immediately decline
by 7.2 % on average in that month; however, revenues will increase by 0.4 % within
6 months of the strike. Revenues in neighboring counties do not experience an immediate change; however, they will eventually rise by 7.9 % on average relative to unaffected
counties.
Stronger hurricanes, on the other hand, have a much more significant impact on tax
revenues. While revenues do appear to recover over time, there is evidence that the
level of taxable revenues in a county hit by a strong hurricane will be substantially
10 Note that these results may be swayed by Hurricane Andrew which only directly impacted MiamiDade and Monroe counties before passing from the Atlantic Ocean into the Gulf of Mexico. The extreme
devastation of Andrew may serve as an outlier for this particular form of analysis.
123
336
A. R. Belasen, C. Dai
lower for at least 6 months as they would have been had the county avoided being
hit by the hurricane. Among directly hit counties, the average impact is an immediate 9.8 % decline. The average aggregate downturn increases roughly 7.4 % after
6 months; however, the evidence points to a resurgence in the latter half of that period.
Stronger hurricanes also cause a downturn in neighboring counties that is similar in
scope to the downturn caused by weaker hurricanes in directly hit counties. Moreover,
the impact of hurricanes is greatest when it first hits land. Among stronger hurricanes,
the first-hit counties face immediate decline in taxable revenues of 33.4 % on average
relative to unaffected counties. Inland counties that are directly hit by the storm experience immediate decreases in taxable revenues of 18.8 % on average. Finally, among
exit counties, the effect diminishes down to just a 9.5 % average decline. Among
weaker hurricanes, there was no statistically significant difference between the three
groups.
Furthermore, we identified that the effect of hurricanes on the overall local economy is different from that on specific subsectors, which indicates that the effect
of hurricanes on a particular county could depend on the economic makeup of the
county, such that this decline may be due to a difference in the economic makeup of
the commercial sector in hurricane-stricken counties. This indicates that the optimal
recovery policy and effort should differ across sectors and therefore, across counties as well, with particular emphasis on recovering potential revenues from tourismbased subsectors since for many counties in Florida, tourism is the largest source
of revenues. We acknowledge that this study and its findings could also potentially
be improved by surveying firms independently of identifying the behavioral choices
made by each firm during the post-hurricane recovery period. Perhaps, this would
reveal more effective techniques for policymakers to follow when pursuing recovery
efforts.
A potential extension would be to examine further subsectors to identify whether
this pattern holds across non-tourist-based subsectors as well. A second extension to
the study would be to examine property damage as property taxes are another key
revenue source for local governments. Finally, a spatial model could be applied to
these data to estimate the decaying impact of the average hurricane as it passes across
Florida. Such a study would be a natural extension to Sect. 6 and could follow the
methodology laid out in Eric Strobl’s (2008) study.
Acknowledgments The authors would like to thank Solomon Polachek, Sajal Lahiri, and seminar participants at SIU Carbondale, along with Jamie Kruse and seminar participants at the 77th Annual Southern
Economic Conference for their comments and suggestions.
Appendix
See Tables 1, 2, 3, 4, 5, and 6.
123
Assessing the impact of hurricanes on localized taxable sales
337
Table 1 Hurricane descriptive statistics
Name
Synoptic
lifecycle
Monetary damage
to FL (in millions
of dollars)
Deaths
in FL
Wind
speed
(mph)
Average
rainfall
Saffir–
Simpson
scale
Andrew
Aug-92
40,500
44
175
5 –7
5
Allison
Jun-95
1.10
0
75
4 –6
1
Erin
Aug-95
0.50
6
87
5 –12
1
Opal
Sep-95
4,600
1
115
5 –10
3
Danny
Jul-97
0
80
2 –7
1
Earl
Sep-98
134 total
to US
104
2
92
6 –16
1
Georges
Sep-98
406.80
0
103
8 –25
2
Irene
Oct-99
1,000
8
75
10 –20
1
Gordon
Sep-00
11.90
1
75
3 –5
1
4
Charley
Aug-04
14,800
29
150
5 –8
Frances
Sep-04
10,200
37
105
10 –20
2
Ivan
Sep-04
9,200
19
130
7 –15
3
Jeanne
Sep-04
7,900
3
121
8 –13
3
Dennis
Jul-05
2,400
14
120
10 –15
3
Katrina
Aug-05
1,700
14
81
5 –15
1
Ophelia
Sep-05
1
80
3 –5
1
Rita
Sep-05
2
115
2 –4
2
Wilma
Oct-05
77.2 total
to US
12,500 total
to US
22,700
31
120
7 –12
3
This table is an updated version of the table found in Belasen and Polachek (2009) using 2010 dollar values.
Some statistics were also taken from Thacker et al. (2008)
Table 2 County-level taxable sales revenue and business density summary statistics
n*t
Overall retail
Number of business units
Number of business units
Revenues (in millions of dollars)
Number of business units
Min
Max
263.52
495.64
6,573.43
10,790.39
−106.18
83
3,950.00
78,415
12,759
Revenues (in millions of dollars)
Bars and taverns
SD
12,795
Revenues (in millions of dollars)
Restaurants
Mean
23.45
42.83
−82.64
344.00
475.53
816.40
4
5,595
11,181
2.95
4.60
84.54
111.37
−27.33
4
32.05
716
The number of observations is measured in terms of counties (n) and months (t). The negative tax revenues
reflect months in which the State of Florida received a net tax loss because of adjustments due to overpayments in previous time periods. While the negative values are a relatively rare phenomenon, the adjustments
occur consistently throughout the data
123
338
A. R. Belasen, C. Dai
Table 3 Fixed effects estimation results of the impact of hurricanes on the natural logarithm of taxable
sales revenue in the average county in Florida
Direct impact
Neighboring impact
Coefficient
Robust SE
Coefficient
Weak hurricane t = 0
−0.0718**
(0.0194)
−0.0264
Weak hurricane t = 1
0.0400*
(0.0207)
Robust SE
(0.0228)
0.0786***
(0.0289)
(0.0227)
Weak hurricane t = 2
−0.0283
(0.0187)
0.0104
Weak hurricane t = 3
−0.0384
(0.0328)
−0.0038
(0.0344)
Weak hurricane t = 4
−0.0582*
(0.0317)
−0.0362
(0.0289)
Weak hurricane t = 5
−0.0521
(0.0327)
−0.0557
(0.0364)
Weak hurricane t = 6
0.0219*
(0.0145)
0.0442
(0.0284)
Strong hurricane t = 0
−0.0981***
(0.0274)
−0.0281
(0.0285)
Strong hurricane t = 1
−0.1095***
(0.0322)
−0.0372
Strong hurricane t = 2
−0.0623
(0.0414)
Strong hurricane t = 3
−0.1056**
(0.0452)
Strong hurricane t = 4
0.0771
(0.0478)
0.0896*
0.0720*
−0.0174
(0.0280)
(0.0369)
(0.0263)
(0.0471)
Strong hurricane t = 5
−0.0319
(0.0553)
−0.0561*
(0.0344)
Strong hurricane t = 6
Month dummies
February
0.1415***
(0.0453)
0.0701*
(0.0399)
0.0390**
(0.0183)
March
−0.0581***
(0.0097)
April
0.0002
(0.0103)
May
0.0997***
(0.0071)
June
0.0347***
(0.0076)
July
0.0074
(0.0103)
August
0.1005***
(0.0132)
September
0.0408***
(0.0123)
October
−0.0439***
(0.0096)
November
0.0582***
(0.0073)
December
0.0941***
(0.0203)
R2
0.6288
F
342.09
n, groups
12,784, 67
Weak hurricanes fall into categories 1, 2, and 3, and strong hurricanes fall into categories 4 and 5
*** Significant at the 1 % level; ** significant at the 5 % level; * significant at the 10 % level
123
Assessing the impact of hurricanes on localized taxable sales
339
Table 4 Fixed effects estimation results of the impact of hurricanes on the natural logarithm of taxable
restaurant sales revenue in the average county in Florida
Direct impact
Neighboring impact
Coefficient
Robust SE
Coefficient
Robust SE
Weak hurricane t = 0
−0.0851**
(0.0401)
0.0049
Weak hurricane t = 1
0.0281
(0.0388)
0.1293
(0.0801)
Weak hurricane t = 2
−0.0611
(0.0393)
0.0288
(0.0740)
(0.0739)
(0.0823)
Weak hurricane t = 3
−0.0168
(0.0782)
0.1383*
Weak hurricane t = 4
−0.1001
(0.0646)
0.0366
(0.0771)
Weak hurricane t = 5
0.0548
(0.0812)
0.0141
(0.0814)
Weak hurricane t = 6
0.0570
(0.0836)
0.0656
(0.0922)
Strong hurricane t = 0
−0.0513
(0.0726)
−0.0342
(0.0639)
Strong hurricane t = 1
0.0505
(0.0896)
−0.0299
(0.0605)
Strong hurricane t = 2
0.0498
(0.0785)
0.0664
(0.0779)
Strong hurricane t = 3
0.0221
(0.0916)
−0.0216
(0.0736)
Strong hurricane t = 4
0.1843
(0.3040)
−0.3847**
(0.1725)
Strong hurricane t = 5
0.0642
(0.0960)
Strong hurricane t = 6
0.2358
(0.2692)
Month dummies
February
0.0586**
(0.0255)
March
−0.0131
(0.0222)
April
0.0953***
(0.0205)
May
0.1746***
(0.0165)
June
0.1144***
(0.0166)
July
0.0453**
(0.0206)
August
0.1661***
(0.0220)
September
0.0566***
(0.0203)
October
−0.0127
(0.0190)
November
0.0828***
(0.0179)
December
0.1101***
(0.0293)
R2
0.2057
F
950.22
n, groups
12,752, 67
0.0071
−0.5315***
(0.0941)
(0.1855)
Weak hurricanes fall into categories 1, 2, and 3, and strong hurricanes fall into categories 4 and 5
*** Significant at the 1 % level; ** significant at the 5 % level; * significant at the 10 % level
123
340
A. R. Belasen, C. Dai
Table 5 Fixed effects estimation results of the impact of hurricanes on the natural logarithm of taxable bar
and tavern sales revenue in the average county in Florida
Direct impact
Coefficient
Neighboring impact
Robust SE
Coefficient
Robust SE
Weak hurricane t = 0
−0.0342
(0.0427)
0.0857
(0.0541)
Weak hurricane t = 1
0.0189
(0.0369)
0.1074**
(0.0510)
Weak hurricane t = 2
−0.0750*
(0.0394)
0.0731
(0.0508)
Weak hurricane t = 3
0.0704
(0.0611)
0.1494*
(0.0778)
Weak hurricane t = 4
−0.0911
(0.0645)
0.0040
(0.0665)
Weak hurricane t = 5
0.0308
(0.0743)
0.0855
(0.0778)
Weak hurricane t = 6
0.0483
(0.0577)
0.0822
(0.0590)
Strong hurricane t = 0
0.1281
(0.0781)
0.0795
(0.0582)
Strong hurricane t = 1
0.0663
(0.0637)
0.0844
(0.0636)
Strong hurricane t = 2
0.1590
(0.1342)
0.1969***
(0.0723)
(0.0690)
Strong hurricane t = 3
−0.0753
(0.1097)
0.0124
Strong hurricane t = 4
0.0007
(0.1451)
−0.2565
(0.1656)
Strong hurricane t = 5
−0.1101
(0.1265)
−0.1129
(0.1044)
Strong hurricane t = 6
Month dummies
February
0.0528
(0.1275)
−0.2307*
(0.1355)
0.0971*
(0.0255)
March
0.0293
(0.0222)
April
0.0688***
(0.0205)
May
0.0816***
(0.0165)
June
0.0655***
(0.0166)
July
−0.0282
(0.0206)
August
0.0584***
(0.0220)
September
−0.0506***
(0.0203)
October
−0.0735***
(0.0190)
November
0.0155
(0.0179)
December
0.1251**
(0.0293)
R2
0.2896
F
n, groups
130.98
11,180, 67
Weak hurricanes fall into categories 1, 2, and 3, and strong hurricanes fall into categories 4 and 5
*** Significant at the 1 % level; ** significant at the 5 % level; * significant at the 10 % level
123
Assessing the impact of hurricanes on localized taxable sales
341
Table 6 Fixed effects estimation results of the impact of hurricanes on the natural logarithm of taxable
sales revenue in the average county in Florida sorted by geographic region
Direct impact
Coefficient
Neighboring impact
Robust SE
Coefficient
Robust SE
Weak hurricanes
H
−0.0455
(0.0351)
−0.0488
(0.0302)
H Entry
−0.1854
(0.0466)
−0.0199
(0.0723)
H Exit
−0.0633
(0.0422)
0.0864
(0.0536)
Strong hurricanes
H
−0.1880***
(0.0367)
−0.0062
(0.0438)
H Entry
−0.1462***
(0.0496)
−0.0804
(0.0743)
H Exit
0.0926*
(0.0515)
−0.0347
(0.0473)
Month dummies
February
0.0460**
(0.0181)
March
−0.0478***
(0.0092)
April
0.0084
(0.0101)
May
0.1074***
(0.0073)
June
0.0425***
(0.0074)
July
0.0156
(0.0102)
August
0.1089***
(0.0132)
September
0.0476***
(0.0121)
October
−0.0296***
(0.0092)
November
0.0629***
(0.0077)
December
0.1024***
(0.0206)
R2
0.6273
F
298.87
n, groups
12,790, 67
Weak hurricanes fall into categories 1, 2, and 3, and strong hurricanes fall into categories 4 and 5
*** Significant at the 1 % level; ** significant at the 5 % level; * significant at the 10 % level
References
Belasen AR, Polachek SW (2008) How hurricanes affect wages and employment in local labor markets.
Am Econ Rev 98(2):49–53
Belasen AR, Polachek SW (2009) How disasters affect local labor markets: the effects of hurricanes in
Florida. J Hum Resour 44(1):251–276
Berry J (2002) Cooking trends in the United States: are we really becoming a fast food country? Energy
Information Administration. http://www.eia.doe.gov/emeu/recs/cookingtrends/cooking.html
Cho S, Gordon P, Moore JE II, Richardson HW, Shinozuka M, Chang SE (2001) Integrating transportation
network and regional economic models to estimate the costs of a large urban earthquake. J Reg Sci
41(1):39–65
Clukey L (2010) Transformative experiences for Hurricanes Katrina and Rita disaster volunteers. Disasters
34(3):644–656
Ewing BT, Kruse JB (2001) Hurricane Bertha and unemployment: a case study of Wilmington, NC. In:
Proceedings of the Americas conference on wind engineering
123
342
A. R. Belasen, C. Dai
Ewing BT, Kruse JB (2002) The impact of project impact on the Wilmington, NC labor market. Public
Financ Rev 30(4):296–309
Ewing BT, Kruse JB, Thompson MA (2005) An empirical examination of the Corpus Christi unemployment
rate and Hurricane Bret. Nat Hazards Rev 6:191–196
Ewing BT, Kruse JB, Thompson MA (2009) Twister! Employment responses to the May 3, 1999, Oklahoma
City Tornado. Appl Econ 41(6):691–702
Guimaraes P, Hefner FL, Woodward DP (1993) Wealth and income effects of natural disasters: an econometric analysis of Hurricane Hugo. Rev Reg Stud 23:97–114
Hallegatte S, Boissonnade A, Schlumberger M, Muir-Wood R (2008) Demand surge and worker migrations
in disaster aftermaths: application to Florida in 2004 and 2005. http://www.centre-cired.fr/IMG/pdf/
Migration_Final.pdf
Kapucu N (2008) Collaborative emergency management: better community organizing, better public preparedness and response. Disasters 32(2):239–262
Kim Y, Kang J (2010) Communication, neighborhood belonging and household hurricane preparedness.
Disasters 34(2):470–488
Rose A, Benavides J, Chang SE, Szczesniak P, Lim D (1997) The regional economic impact of an earthquake:
direct and indirect effects of electricity lifeline disruptions. J Reg Sci 37(3):437–458
Rose A, Liao S (2005) Modeling regional economic resilience to disasters: a computable general equilibrium
analysis of water service disruptions. J Reg Sci 45(1):75–112
Strobl E (2008) The economic growth impact of hurricanes: evidence from US coastal counties. IZA
Discussion Paper No. 3619
Thacker MTF, Lee R, Sabogal RI, Henderson A (2008) Overview of deaths associated with natural events,
United States, 1979–2004. Disasters 32(2):303–315
West C, Lenze D (1994) Modeling the regional impact of natural disaster and recovery. Int Reg Sci Rev
17(2):121–150
Xiao Y (2011) Local economic impacts of natural disasters. J Reg Sci 51(4):804–820
Xiao Y, Feser E (2013) The unemployment impact of the 1993 US midwest flood: a quasi-experimental
structural break point analysis. Environ Hazards. doi:10.1080/17477891.2013.777892
123
Copyright of Annals of Regional Science is the property of Springer Science & Business
Media B.V. and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.
How Disasters Affect Local Labor Markets: The Effects of Hurricanes in Florida
Author(s): Ariel R. Belasen and Solomon W. Polachek
Source: The Journal of Human Resources, Vol. 44, No. 1 (Winter, 2009), pp. 251-276
Published by: University of Wisconsin Press
Stable URL: http://www.jstor.org/stable/20648894
Accessed: 08-10-2017 03:01 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
University of Wisconsin Press is collaborating with JSTOR to digitize, preserve and extend
access to The Journal of Human Resources
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
How Disasters Affect Local Labor
Markets
The Effects of Hurricanes in Florida
Ariel R. Belasen
Solomon W. Polachek
ABSTRACT
This study improves upon the Difference in Difference approach by
examining exogenous shocks using a Generalized Difference in Differenc
(GDD) technique that identifies economic effects of hurricanes. Based on t
Quarterly Census of Employment and Wages data, worker earnings in
Florida counties hit by a hurricane increase up to 4 percent, whereas
earnings in neighboring counties decrease. Over time, workers experienc
faster earnings and slower employment growth than workers in unaffect
counties. Hurricanes have a greater impact in coastal and Panhandle
counties, and powerful hurricanes have greater economic effects than weake
ones. Further, the GDD technique is applicable to analyze a wider range
exogenous shocks than hurricanes.
I. Introduction
An exogenous shock is an unexpected event that impacts a given
market. Such shocks can take many forms, ranging from unexpected new legislation,
to sudden population shifts, to domestic weather-related events, and even to terrorist
attacks. A number of studies utilize Difference-in-Difference (DD) estimation to ex
amine the effects of exogenous shocks. For example, Card (1990) in a well-cited ar
ticle used DD to examine migration and found relatively small effects on wages.
Such studies look at changes across time periods between the region of interest
and a comparable region which was unaffected by the shock to find long-run effects.
Angrist and Krueger (1999) call these results into question for failing to identify an
Ariel R. Belasen is an assistant professor of economics and finance at Southern Illinois University in
Edwardsville, III. Solomon W. Polachek is a distinguished professor of economics at the State University of
New York at Binghamton. All data contained in this paper will be made available upon request. The authors
would like to thank Joel Elvery, Christopher Hanes, Kajal Lahiri, and Stan Masters along with the seminar
participants at: SUNY Albany; SUNY Binghamton; the Third Migrant ITA Conference, Washington, D.C.
March 2007; and the Twelfth Society of Labor Economists Meetings, Chicago, May 2007for their comments.
[Submitted July 2006; accepted September 2007]
ISSN 022 166X E ISSN 1548 8004 ? 2009 by the Board of Regents of the University of Wisconsin System
THE JOURNAL OF HUMAN RESOURCES ? 44 ? 1
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
252 The Journal of Human Resources
appropriate control group. Perhaps, as a result, there is now a literature on appropri
ately choosing control groups, for example Bertrand, Duflo, and Mullainathan
(2002), Kubik and Moran (2003), and Abadie, Diamond, and Hainvellen (2007). An
other problem is the experimental group. Most papers examining exogenous shocks
rely on one experimental group; in Card's (1990) case, this experimental group is
Miami, the site of the Mariel Boatlift. However it is not obvious that one experimen
tal group suffices. In the Card example, the Miami labor market might not be typical
of other potential experimental sites. Perhaps in his study, Miami's unemployment
did not rise because Miami's economy was growing more rapidly than other simi
larly sized cities. This paper finds that by properly addressing these two issues,
one can better isolate the direct impact of exogenous shocks. We find that counties
hit by hurricanes experience a positive net effect on earnings and a negative net effect
on employment, but that these effects dissipate over time.
One innovation of this paper is to have many random experimental groups as well
as many random control sites. To achieve this, we use a different natural experiment,
hurricanes, to examine the effect of an exogenous shock on a local labor market.
Hurricanes, in particular, are a good choice for this study because they can affect sev
eral counties at a time, and can occur more than once in the time period under study.
By having many experiment sites, we are able to test how the impact of exogenous
shocks differs by both characteristics of the shock and characteristics of the experi
ment group. Other papers have used weather-related events such as rainfall (includ
ing Miguel 2005; Waldman, Nicholson, and Adilov 2006; and Connolly 2007) to
obtain a purely exogenous variable as an instrument to predict other independent var
iables such as how much television children watch (in the case of Waldman et al.
2006), which in turn is used to predict autism using a simultaneous equation ap
proach. We use weather (that is hurricanes) directly as the exogenous shock we want
to evaluate.
To do this, we develop a Generalized Difference-in-Difference (GDD) technique
in which we compare affected regions to unaffected regions across multiple exoge
nous events and time periods. In addition, exogenous shocks that are felt positively
by one specific labor market can also have an effect on nearby labor markets. Thus
we can examine multiple exogenous shocks affecting more than one locality at a
time. Further, to address the issue of the appropriate definition of treatment and con
trol groups, we compare a given hurricane-stricken county to all other unaffected
counties within that state. In addition, by using quarterly time-series data, this ap
proach has the advantage of distinguishing short-term and long-term effects that pre
viously had been neglected. In this way we can better identify the effect of an
exogenous shock as well as quantify its effect over time.
The destructive power of hurricanes worldwide can wipe out thousands of lives
and cause billions of dollars worth of infrastructure and private property losses an
nually. Hurricane season runs from June 1st through November 30th each year over
warm water, defined as oceanic temperatures exceeding 80 degrees Fahrenheit. How
ever, the exact timing and path of the hurricanes cannot be determined in advance.
Due to the high temperatures required, most hurricanes that strike the United States
strike the Gulf States and the Southeastern States. Because Florida is a member of
both subsets of states, it is instructive to look at the county-level Florida labor market
to examine the exogenous shocks of hurricanes.
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
Belasen and Polachek
Over the course of an average year, the state of Florida will generally see one to
two hurricanes during the six-month hurricane season, but there are years when Flor
ida is not hit, even once. Over the last two years of the sample (2004 and 2005), how
ever, the hurricanes that struck Florida were more frequent and more powerful than
ever before.1 Although hurricanes are not completely unexpected shocks to the state
of Florida, each hurricane event is exogenous to the specific counties that are hit as
well as to the degree of damage unleashed. Therefore, the events we have identified
can be used as an independent variable by comparing those counties that have been
hit to the other counties that avoided devastation.
Florida is comprised of 67 counties and, over the past 18 years, none of them have
escaped the effects of hurricanes. Five of the six most damaging Atlantic hurricanes
of all time have struck Florida over the course of this time period. Damages to prop
erty can be estimated in direct monetary costs, for example, 1992's Hurricane
Andrew wound up costing Southern Florida roughly $25.5 billion ($43 billion in
2005 USD) in property losses (Rappaport 1993). However, a county, business or per
son's wealth is made up of more than just the stock of assets owned by that person. A
major portion of the flow of one's wealth comes from earned income. Thus the ques
tion is raised, how can the income-specific and employment-specific effects of a hur
ricane be measured? In addition, when looking at the effects of a hurricane on a
specific county, are there any spillovers that need to be accounted for in neighboring
counties? In addition, do more destructive hurricanes impact labor markets more
intensely? And finally, how long are the effects of a hurricane felt in earnings and
employment?
II. Background on Florida and
the Hurricanes
Over the course of the last 18 years, the state of Florida has been rav
aged by 19 hurricanes. A summary table containing descriptive statistics for each of
the hurricanes can be seen in Table 1, which lists magnitude, monetary costs, and
death statistics for each storm. Each hurricane is given a standard name by the World
Meteorological Organization assigned to the storm in alphabetical order each year
based on the timing of the storm. The lists of names for hurricanes change each year,
with the gender of the initial storm also alternating each year. There are six lists in
total and any time a particularly devastating hurricane occurs, the name of that hur
ricane is "retired" from the list (Padgett, Beven, and Free 2004). After the sixth list
is used, the first is then cycled back with any retired hurricane names replaced with
new names beginning with the same letter as the retired ones.
1. The National Oceanic and Atmospheric Administration retires the names of particularly devastating hur
ricanes. Nine of the nineteen hurricanes in the sample occurred in the 2004 and 2005 hurricane seasons.
Eight of those storms have had their names retired (as opposed to just three retirees throughout the remain
der of the sample), including Hurricane Wilma which set records for intensity. Note, however, that in this
past 2006 season, Florida was only hit by one minor hurricane: Ernesto, so this is not necessarily a trend
moving forward.
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
254 The Journal of Human Resources
U
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
Belasen and Polachek
Hurricanes are categorized according to the Saffir-Simpson Scale based on their
wind speed. Hurricanes Florence, Allison, Erin, Danny, Earl, Irene, Gordon, Ophelia,
and even the Floridian part of Katrina were Category 1 hurricanes at landfall, mean
ing they had wind speeds ranging between 74 and 95 miles per hour. Hurricanes
Georges, Frances, and Rita were Category 2 hurricanes and had wind speeds ranging
between 96 and 110 miles per hour. With wind speeds ranging between 111 and 130
miles per hour, Hurricanes Opal, Ivan, Jeanne, and Dennis were classified as Cate
gory 3 hurricanes. Hurricane Charley reached 150 miles per hour and became Cat
egory 4 as it hit the mainland. Hurricanes Andrew and Wilma were Category 5
hurricanes and had winds well above 180 miles per hour.
III. Economic Model of Hurricanes
According to Lucas and Rapping (1969), when people perceive a
shock as having a temporary effect on the economy, they will not alter their long
term perception of the economic variables that are affected by the shock. Hurricanes
generally last for, at most, two or three days once they strike land. Historically speak
ing, even the damages from the most destructive hurricanes are typically repaired
within two years of the hurricane. Therefore, one would expect to see perceptions
of the future remain largely unchanged in the long run as the variables return to their
steady state levels of growth. Guimaraes, Hefner, and Woodward (1993) state that
while hurricanes create an economic disturbance in the short run, oftentimes they
can lead to economic gains in the long run.
More specifically, within labor demand and labor supply, hurricanes will lead to
negative shocks on labor supply in the stricken region, along with undetermined
shocks to the region's labor demand as some firms attempt to fill vacancies in their
work force while others leave town with the outflow of workers. If a hurricane strikes
a region and causes people to flee, the work force in that region will decrease. There
fore, labor supply would shift downward. At the same time, if that hurricane destroys
a lot of private property and physical capital, labor demand could also decrease as
employers have to close their shops. However, Skidmore and Toya (2002) point
out that the risk of a natural disaster can reduce the expected return to physical cap
ital (which may be destroyed during the storm) and, in turn, there is a substitution
effect toward human capital as a replacement. Of course, as the demand for human
capital rises, the price of human capital will also rise. This leads to an income effect
that runs counter to the substitution effect. On the other hand, if the hurricane only
destroys residential areas, labor demand also could increase as employers attempt to
fill vacant jobs. Thus, the shock on labor demand from a hurricane most likely will be
positive leading to changes in earnings and employment.
Using the standard labor market framework, with labor supply shocked negatively
and labor demand shocked positively, earnings will increase, and employment will
have an ambiguous effect depending on whether or not the demand shock outweighs
the supply shock. The set of earnings and employment that we are examining in this
study are county-level average quarterly earnings per worker in the state of Florida.
In order to measure the actual earnings effects of hurricanes on earnings, we will
control for other factors that have an effect on earnings and employment. Florida's
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
The Journal of Human Resources
economy has been growing rapidly over the last half-century and every county in
Florida has benefited from this growth. Card (1990) found that immigrants in Miami
had no long-term effects on wages despite increasing the labor force by 7 percent. He
deduces that the Florida labor market in the 1980s was able to simply absorb a group
of 45,000 immigrants into the labor market without a change in wages because of the
rapid growth of Florida's economy. E wing and Kruse (2005) isolated the specific
county-level fluctuations from the overall general growth by controlling for the trend
of earnings movement across the entire state. In a subsequent paper, Ewing, Kruse,
and Thompson (2007) explained that local economies may be influenced by state
business cycles. Following their method, we control for the state trends of Florida.
Furthermore, Florida's labor market is greatly influenced by seasonal shifts. During
the summer months, earnings and employment decrease in several sectors of the la
bor market. Thus, one must also control for seasonality.
In the end, we have two equations, one for employment (Qit) and one for earnings
(yit) which sets the dependent variable equal to a function of state (Qt, yt), county-spe
cific time-invariant effects (Z,), seasonal trends (St) as well as hurricane effects (Hit):
(1) Qit=f{Qt,ZhSt,Hit) + uit
(2) yit=f(ytlZhShHit) + vit
As stated earlier, an important question to consider when examining hurricanes
and other exogenous shocks is what kind of neighboring effects, if any, will affect
the model. If a hurricane forces workers to flee one county for a second county, then
labor supply in the original county will be negatively affected while labor supply in
the second county will be positively affected. Thus, the model is set up to include a
series of hurricane dummy variables that capture direct effects and neighboring
effects. This allows us to compare three distinct sets of counties: those that were di
rectly hit and faced heavy destruction, those that were close by, and thus affected by
heavy rainfall, and those that were farther out, and generally unaffected by the hur
ricane. Assuming that counties i and j border one another, the subscript i under HD
indicates that the locus of destruction2 from the hurricane is directly passing over
county i while subscript ij under HN indicates that the locus of destruction of a hur
ricane is passing through county j which borders county /. In other words, HD takes a
value of one when the hurricane strikes county /; and HN takes a value of one when
the hurricane strikes county j but not county /. More specifically,
(3) Qit = %uQt + 2/ ; + 3 ?3 + 4 + 5/ ? + %iHl + uit
(4) yu =
a, 5
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
260 The Journal of Human Resources
o
S
i
r
o
.O
a
3
?2
1I
'C
?
a
: 5 "O
U o.
ifi
a.
This content downloaded from 131.247.112.3 on Sun, 08 Oct 2017 03:01:33 UTC
All use subject to http://about.jstor.org/terms
? ? .s
Belasen and Polachek 261
m os
m o ?s
O
O O S
'
Purchase answer to see full
attachment