Transcript ESTIMATING THE ECONOMIC IMPACT OF TERRORISM:

Discussion on Extended Validity of an
Alternative Framework to Estimate
Short-Term Negative Impacts of an
Unexpected (Unprecedented) Event
The 5th Business Enterprises for Sustainable
Travel Education Network Think Tank and
Conference
Session 2B: Economics of Tourism Crises and
Disasters 11:45~ June 17th 2005
Tad Hara: (Presenter) Regional Science Program &
School of Hotel Administration, Cornell University
Prologue: Economic Impact of Terrorism

Negative Impact of Terrorism over Economy


Widely assumed, but hardly quantified
Vulnerability of Certain Sectors

E.g. Tourism industry


One of the largest & fastest growing export sectors for many
nations, its potential looks bright, but its vulnerability (risk of
revenue volatility) seldom quantified
Benefit of quantification for policy makers

Proper response, better plan for prevention


June 17th, 2005
E.g. Which industry will suffer most, jobs, income, tax revenue,
etc.
How much budget can be justified for preventive measures
BEST-EN by Tad Hara
2
Tourism Industry
“Tourism can be one of the
few development opportunities
for the poor.” World Tourism
Organization 2002
“Denpasar, Bali, Indonesia, Sunday,
Oct. 13, 2002. (AP Photo/Jack
Hamilton)
June 17th, 2005
BEST-EN by Tad Hara
3
Prologue: Economic Impact of Terrorism
“Facts About Bali: ECONOMY: Known
as the Island of the Gods, Bali has
been one of the world's most popular
tourist destination for decades. The
likelihood of the blast scaring
away hundreds of thousands of
tourists could be a devastating
blow to not only the island's
economy but that of the whole
of Indonesia, which is struggling to
recover from the Asian financial crisis
of the late 1990s.
Sat Oct 12,10:33 PM ETBy The
Associated Press
Need to study this topic?
June 17th, 2005
BEST-EN by Tad Hara
4
Economic Impact of Terrorism: Introduction
Economic Impact of Terrorism is hard to
quantify – Best estimates of “experts” often
used
 How do we estimate?

1 Quote whatever the other sources said
 2 Econometrics / Time Series Model
 3 Input Output/Social Accounts framework

June 17th, 2005
BEST-EN by Tad Hara
5
Literature Review (1/5)
 Literature
on the Negative Economic
Impact of Natural Disasters
– Horwich (2000): “it is capital stock, not output,
that is directly reduced by the disaster” (Kobe
Earthquake)
– Murdoch, Singh & Thayer: S.F Earthquake
caused 2% reduction in housing values.
June 17th, 2005
BEST-EN by Tad Hara
6
Literature Review (2/5)

Negative Economic Impact of a Man-Made
Disaster
– Cohen (1995) Impact of Exxon Valdez Oil Spill
 Concede difficulty of Ex-post analysis
– Globar (1993) Econometric model (ex-post
analysis)

Identified Inverse relation between defense expenditure
and investment in Sri Lanka.
– Coshall (2003) Intervention analysis (impacts of
events on UK Air travel using ARIMA)
– Negative Surprise tend to increase volatility than
positive one (asymmetry of volatility: Engle et al
1990)
June 17th, 2005
BEST-EN by Tad Hara
7
Literature Review (3/5)

Input Output Models
– Isard & Kuenne (1953)


Estimating Impact of Steel industry in NY-Philadelphia region
Demonstrated employer effect is much higher than prior
assumption of 1:1.
– Ahlert (2001)
 Impact of Soccer World Cup 2006 to German Economy
– English (2002)
 Impact of Converting Corn to Ethanol Production
– Nakajima (1994)
 International I-O model (Japan & Asia)
– Lee (1994)
 I-O with fuzzy final demands
June 17th, 2005
BEST-EN by Tad Hara
8
Literature Review (4/5)

I-O Model Application to Tourism Industry
– Fletcher (1981)


Economic Impacts on Gibraltar
While final demand is dominated by UK defense
expenditure, tourism generated the highest marginal
increase in income and employment
– Archer (1982)
 Usage of I-O for tourism policy
– Heng & Low (1990)
 Tourism industry’s impact on Singapore Economy
June 17th, 2005
BEST-EN by Tad Hara
9
Literature Review (5/5)

Negative Impacts with the I-O model
– Caskie, Davis & Moss (1999)
 Simulation of Negative Impacts of BSE on N. Irish
Economy
– Zhou, Yanagida, Chakravorty & Leung (1997)
 Simulation of 10% decrease in tourism revenue in
HI (partially using CGE by GAMS)
– Okuyama, Hewings & Sonis (1997)
 Kobe Earthquake’s shockImpacts from
unscheduled events are not only the negative effects
but also positive effects of reconstruction.
June 17th, 2005
BEST-EN by Tad Hara
10
Estimation--Assumptions

Initial Shock – fairly simple one
– Assume that the reductions in number of employed
between Sept. and Oct. 2001 are all attributed to the
huge exogenous shock
– Allocate the initial shocks to SIC 1-digit industry
categories
June 17th, 2005
BEST-EN by Tad Hara
11
Simulations—Initial Shock
(Services adjusted: Adjusted Version)
Table 4-7:The Initial Shock of 9/11 to Jobs in NY City and NY State (Service Adjusted)
Industrial Sector
Mining
Construction
Manufacturing
TCPU
Trade
FIRE
Services
Total
NY CITY
0
+700
-4,300
-4,300
-9,900
-25,400
-1,900
-45,100
Industrial Sector
Mining
Construction
Manufacturing
TCPU
Trade
FIRE
Services
Total
NY STATE
+100
-1,800
-7,200
-5,500
-6,800
-25,000
-5,700
-45,780
Source: The authors based on data from New York State Department of Labor, 2002
June 17th, 2005
BEST-EN by Tad Hara
12
Simulations--Results
(Services Adjusted: NYS Results: Latest Version)
Table 4-8: Summary of the 9/11 WTC attack Impact simulation for New York STATE
(Latest Version & Services sector Impact adjusted)
Impact on NY State Direct
(the initial
impact by 9/11)
Total Value Added
Total Output
Employment
Indirect
-$2,136M
-$8,066M
-$3,479M
-$11,712M
-48,727 jobs -25,651 jobs
Induced
-$2,627M
-$4,057M
-38,856 jobs
Total Effect
-$12,830M
-$19,248M
-113,234 jobs
Source: The authors in August 2002 using IMPLAN. The initial shocks to the model are based on
revised data from the New York State Department of Labor
June 17th, 2005
BEST-EN by Tad Hara
13
Simulations--Results
(Services Adjusted: NYC results: Adjusted version)
Table 4-9: Summary of the 9/11 WTC attack Impact simulation for New York City (Latest
Version & Services sector Impact adjusted)
Impact on NY City
Total Value Added
Total Output
Employment
Direct
(the initial
impact by 9/11)
Indirect
-$1,882M
-$9,726M
-$2,692M
-$13,283M
-42,780 jobs -15,216 jobs
Induced
-$1,398M
-$1,976M
-15,430 jobs
Total Effect
-$13,006M
-$17,952M
-73,427 jobs
Source: The authors using IMPLAN. The initial shocks to the model based on data from New
York State Department of Labor
Thus, the revised initial impact estimation without the
effects of growth of the irrelevant sector (education) can
change the resulting total impact significantly.
June 17th, 2005
BEST-EN by Tad Hara
14
Review of Other Studies

NY City Economic Impact  NY State Jobs Lost
– “Total Loss $83
billion” (NYC
partnership & Chamber of
Commerce: Nov 2001)
– “Total Cost $54 billion” (NY
Governor: Oct 2001)
– “WTC Replacement Cost & Cleanup
$25~29 billion” (FEB NY: April
2002)
– “Total Cost $83 billion (quoting
NYCP-COC) but $67 billion covered
by Insurance (US GAO: May 2002)

– “99,000 in 2001, 78,000 in
2002, 77,000 in 2003” (NYS
Senate Finance Committee:
DRI-WEFA: January 2002)
– “Resulted at peak loss of
78,200” (DRI-WEFA: March
2002)
– “50,000 immediately, 70,000
in 4th Quarter” “Much of this
loss is likely linked to WTC
attack” (FEB NY: April 2002)
NY City JOBS Lost
– 108,500, 115,300, 105,200,
125,000, 84,000, 78,200,
129,000….
June 17th, 2005
BEST-EN by Tad Hara
15
Ch 4: Verdict--Forecast VS Reality: NYC
Change of Employment in NYC
Our Forecast
-73,400 for
NYC
AVERAGE
2002-Sep
-77.4
-87
2002-Aug
-103.1
2002-Jul
-99.7
2002-Jun
-65.4
2002-May
Month
-77.8
-89.3
2002-Apr
NY City
2002-Mar
-103.7
-119.7
2002-Feb
-122
2002-Jan
-12.9
2001-Dec
-18.2
2001-Nov
-30.5
-140
-120
-100
-80
-60
-40
2001-Oct
-20
0
Change in # of Jobs in thoudands
June 17th, 2005
BEST-EN by Tad Hara
16
Ch 4: Verdict--Forecast VS Reality: NYS
Change of Employment in NYS and NYC
AVERAGE
Our Forecast
-113,200 for
NYS
-115.9
2002-Sep
-87.1
2002-Aug
-90.3
2002-Jul
-91.3
2002-Jun
-56.8
2002-May
Month
-103.4
-172.5
2002-Apr
NY STATE
2002-Mar
-218.1
-257.8
2002-Feb
-266.2
2002-Jan
-5.9
2001-Dec
-15.9
2001-Nov
-25.6
-300
-250
-200
-150
-100
-50
2001-Oct
0
Change in # of Jobs in thousands
June 17th, 2005
BEST-EN by Tad Hara
17
Ch 4: Verdict--All Studies In Perspective
Compare with Actual Data
Table 12: Changes of employment in NYC and NYS since September 2001
Number
change from
Number
change from
NYC
NYS
of jobs
2001
2001
2001
2001
2002
2002
2002
2002
2002
2002
2002
2002
2002
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Sept.
of jobs
3,136.5 (in thousands)
3,106.0
(30.5)
3,118.3
(18.2)
3,123.6
(12.9)
3,014.5
(122.0)
3,016.8
(119.7)
3,032.8
(103.7)
3,047.2
(89.3)
3,058.7
(77.8)
3,071.1
(65.4)
3,036.8
3,033.4
3,049.5
2001
2001
2001
2001
2002
2002
2002
2002
2002
2002
(99.7)
2002
(103.1)
2002
(87.0) (77.4) 2002
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Sept.
7,186.3 (in thousands)
7,160.7
(25.6)
7,170.4
(15.9)
7,180.4
(5.9)
6,920.1
(266.2)
6,928.5
(257.8)
6,968.2
(218.1)
7,013.8
(172.5)
7,082.9
(103.4)
7,129.5
(56.8)
7,095.0
7,096.0
Not yet
(91.3)
(90.3)
(87.1) (115.9)
Source: The authors based on data from New York State Department of Labor, 2002
June 17th, 2005
BEST-EN by Tad Hara
18
Ch 4: Verdict—Historical Data
Change in Jobs in
'000
NYC Monthly Employment Changes:
Averages since 1976-77 vs 2001-02
60
40
20
0
-20
-40
-60
-80
-100
-120
-140
9-10
1011
1112
12-1
1-2
2-3
3-4
4-5
5-6
6-7
7-8
8-9
Monthly Change (Sep/Oct to Aug/Sep)
June 17th, 2005
BEST-EN by Tad Hara
Average
Excl.2001-02
2001
19
Concluding Remarks & Future Research

Concerns
Important assumptions
 Varieties of interesting problems in “ex post”
analysis
 Time Frame Issues (in what time span)
 Rooms for Refinement

Econometric Model Let’s have a
quick look
 Time Series Analysis
 Multivariate
June 17th, 2005
BEST-EN by Tad Hara
20
Review of Other Types of Models

1. Multivariate Econometrics Model
yi   0  1 X 1   2 X 2  ...   i
– Future depends on past associations of dependent variable with independent data
– Convenient model, widely used, convincing
 Works well after all the data are on the table (when dust settles down)

2. Time Series Model
– Future depends on past behaviors of the own data
– Powerful, good at finding patterns
yt  1 yt 1  2 yt 2  ... p yt  p  wt
– But…“Because many models make predictions based on
market’s past behavior, they can be easily wrong-footed by
unusual market moves”
– “We look at the worst probable risk, not the worst possible risk.”
– WSJ9/27/02 “Rocky Markets Foil Firms’ Bets Based on ‘Risk Models’
June 17th, 2005
BEST-EN by Tad Hara
21
Review of Econometrics model : Example
the "labor endowment" (working-age population) of the state economy E;
state after-tax unearned income Yu;
state government transfer payments Gtr;
the federal individual income tax rate tf;
the state individual income tax rate ts;
the U.S. unemployment rate u;
the state unemployment insurance tax rate v Ð unemployment insurance
contributions divided by payroll;
the components of the cost of capital to firms, the discount (interest) rate i,
the capital replacement rate d, the present value of depreciation allowed for
tax return purposes for $1 of capital c, and the total tax rate on corporations
(including tax on dividends, capital gains, and income) r.
the Massachusetts workers compensation tax rate wc.
Can you find these
data at t=0? Or
How long you wait?
lnL = 2.3577lnE - 0.0902lnGtr + 0.0698lnYu 0.005tf - 0.033 ts - 0.011u - .043v + 0.003 d +
0.002c + 0.004i - 0.001r - 0.022wc (10.472)* (-1.845) (2.61)*
(-1.9) (-2.93)* (-4.83)* (2.36)* (-2.88)* (1.31)* (1.28) (-.87) (-1.81)
Source: Kathleen M. Lang, PhD, BHI research associate, and Howard R. Wright, BS, BHI research economist,
prepared this BHI FaxSheet
http://www.beaconhill.org/FaxSheets/FaxUIFeb97.html
June 17th, 2005
BEST-EN by Tad Hara
22
Review: I-O/SAM model
1
X  ( I  A)  Y
Where, X=Total Output (an nx1 vector), I=Identity
Matrix an nxn matrix), A=“A” matrix (standardized
inter-industry coefficient matrix: nxn), Y=Final
demand (an nx1 vector)
You may rewrite the equation….
Y  ( I  A)  X
Increase [loss] in final demand in some sector will result in
Increase [loss] in Total Output (overall economic activities)
June 17th, 2005
BEST-EN by Tad Hara
23
Data Availability?
How many
months? How
many years
later?
Source: Time managize: Suzanne Plunkett/AP http://www.time.com/time/photoessays/wtc/7.html
June 17th, 2005
BEST-EN by Tad Hara
24
Review of Time Series model: Example
Time Series Analysis
e mp r a t e
7400
7300
For ec as t
f or
7200
e mp r a t e
7400
7300
7200
7100
7000
6900
6800
6700
7100
7000
6900
6800
6700
6600
6500
6400
6300
6200
6100
6000
5900
5800
5700
5600
5500
5400
5300
6600
6500
6400
6300
6200
6100
6000
5900
5800
5700
0
5600
100
5500
200
300
400
t
5400
5300
For ec as t
J A N1 9 7 5
J A N1 9 8 0
J A N1 9 8 5
J A N1 9 9 0
J A N1 9 9 5
J A N2 0 0 0
J A N2 0 0 5
f or
e mp r a t e
7300
7200
7100
e mp d a t e
7000
6900
6800
6700
6600
6500
Source: the authors by SAS based on data from Labor
Department of New York State.
6400
6300
6200
6100
6000
5900
5800
5700
5600
5500
5400
5300
June 17th, 2005
BEST-EN by Tad Hara
0
100
200
t
25
300
400
Concluding Remarks & Future Research

Advantages of I-O/SAM framework for
Estimating Negative Impacts of Terrorism




General Criticism of I-O/SAM: “No Capacity
Constraints” does not hold
Structurally good at capturing “a huge shock”
Superior to other models which may estimate
direct impacts only in that I-O/SAM will capture
total negative impacts to a national/ regional
economy by describing indirect and induced
effects of terrorism.
External Validity

June 17th, 2005
We can quantify impacts of large negative shock for a
nation/region GIVEN I-O/SAM table and solid labor data.
BEST-EN by Tad Hara
26
E ( i2 )   2
Concluding Remarks & Future Research
 Rooms
for Improvements
 General
EquilibriumCGE model (you can make variables
endogenous; interest rate, price, exchange rate etc:
WARNING: CGE’s accuracy highly depends on the very
accuracy of I-O/SAM data)
 Hybrid
with econometrics for final demand (∆FD), some
stochastic model
 (Once data become available), ARCH, GARCH, EGARCH, or Structural Equation with Regime Shifts?


Can We Predict whether structural shift occurred given current
surge/plunge of data (t=0)? Frontier topic of Time Series
Based on constraints of yt = αyt-1 + εt .
it is challenging to endogenize the shock. (Bayesian, Kalman
filter?)
 I-O/SAM
seem to have good validity to deal with huge
exogenous shocks amid the chaos immediately after the
June 17th, 2005
by Tad Hara
27
huge unprecedentedBEST-EN
shock.
Concluding Remarks & Future Research

Research on Negative Shocks: Economic
Impacts and Beyond
 Quantification
of potential threatsEasier to justify
expenditures for proactive measures
 Final demandinfluenced by psychological factors?
 Roles of mass media--intermediary of information

The whole system of Terrorism
 Confrontational
eradication by “War on Terrorism”, or
 Research on why some resort to terrorism?—
multidisciplinary subjectits understanding may provide
cost-effective prevention

Systematic Prevention is important to Realize
Potential of Stable Growth of Tourism Industry
June 17th, 2005
BEST-EN by Tad Hara
28
Case: WTC Attacks
on 9/11/2001 NY City
June 17th, 2005
BEST-EN by Tad Hara
29
Geographical Distribution of Terrorism Events
Will you believe…that history repeats itself?
Figure June
1-1 17th,
"Patterns
of Global Terrorism:
2000."
2005
BEST-EN
by Tad Hara
Source: U.S.State Departmenthttp://www.state.gov/s/ct/rls/pgtrpt/2000/
30