TERRORISM?SEFFECTONTOURISM:DEVELOPED VS.DEVELOPING
COUNTRIES
Exceptwherereferenceismadetotheworkofothers,theworkdescribedinthisthesisis
myownor wasdoneincollaborationwithmyadvisorycommittee.Thisthesisdoesnot
includeproprietaryorclassifiedinformation.
________________________
AlexiSimosThompson
Certificateof Approval:
________________________________________________
RichardP.Saba JohnD.Jackson,Chair
Professor Professor
Economics conomics
_______________________________________________
T.RandolphBeardGeorgeT.Flowers
Professor InterimDean
Economics raduateSchool

TERRORISM?SEFFECTONTOURISM:DEVELOPED VS.DEVELOPING
COUNTRIES
AlexiSimosThompson
AThesis
Submittedto
theGraduateFacultyof
AuburnUniversity
inPartialFulfillmentofthe
Requirementsforthe
Degreeof
MasterofScience
Auburn,Alabama
August9,2008

iii
TERRORISM?SEFFECTONTOURISM:DEVELOPED VS.DEVELOPING
COUNTRIES
PermissionisgrantedtoAuburnUniversitytomakecopiesofthisdissertationatits
discretion,uponrequest ofindividualsorinstitutionsandattheirexpense.Theauthor
reservesallpublicationrights.
_____________________
SignatureofAuthor
_____________________
DateofGraduation

iv
THESISABSTRACT
TERRORISM?SEFFECTONTOURISM:DEVELOPED VSDEVELOPING
COUNTRIES
AlexiSimosThompson
MasterofScience,August9,2008
(B.S.,AuburnUniversity,2003)
70TypedPages
DirectedbyJohnD.Jackson
Thisthesisstudiesthe effectterrorismhasontourismbetweendeveloping and
developedcountries.Usingcrosssectionaltechniquesonadatasetconsistingof60
countries,resultsconcludethatterroristeventsare moredamagingtodevelopingcountry
tourismthantodevelopedcountrytouristsectors.Possibleexplanationsforthe
difference are givenaswellaspolicy adjustmentsbasedontheresults.
Theorganizationofthepaperisasfollows. Chapter1includesabriefhistoryof
tourismandterrorismandhowthey affecteachother.Chapter2isaliteraturereview
includingadiscussionof thetourismmodel.Chapter3presents methodology.The
modelinthisthesisisdiscussedinthischapter aswellasexplanationofallvariables.
Chapter4discussestheresultsoftheregressionsandothertestsconductedonthedata

v
includingmisspecificationtests.Finally,Chapter 5consistsoftheconclusionwhich
discussespolicysuggestionsbasedontheresultsofthisthesis.

vi
Stylemanualorjournalused:AmericanEconomicReview
Computersoftwareused: Limdep
Excel2007

vii
TABLEOFCONTENTS
LISTOFTABLES x
LISTOF FIGURES xi
1.INTRODUCTION 1
A.LookatTourism 1
B.A LookatTerrorism 4
C.Terrorism?sEffectonTourism 5
D.ConcludingRemarks 6
E.StudyOverview 7
2. LITERATUREREVIEW 8
A.Previous Literature 8
B.Conclusion 13
1. METHODOLOGY 14
A.Model 14
B.Data 20
C.DescriptiveStatistics 21
D.Jarque-Bera NormalityTest 32
2. RESULTS 33
A.Introduction 33

viii
B.Tables andResults 33
C.MoreEvidence ofDifferencesBetweenDeveloped&Developing
Countries 38
D.Testing forHeteroskedasticity 41
E.Misspecification 42
F.ModelComparison 43
G.TourismExpenditureModel 44
a. BreuschPagan-Model2 46
b. MisspecificationModel 2 46
3. CONCLUSION 49
4. SUMMARY 52
REFERENCES 53
REGRESSIONANALYSISRESULTSUSING LIMDEP
Model1 55
Model2 56
Model3 57
APPENDIX 58

ix
LISTOFTABLES
Table1TourismGrowthSince1950 2
Table2DescriptiveStatisticsofTPC(innat.logform) 23
Table3DescriptiveStatisticsofTERR(innat.logform) 25
Table4DescriptiveStatisticsofPPP(innat.logform) 27
Table5DescriptiveStatisticsofGDPS(innat.logform) 29
Table6DescriptiveStatisticsofGDPH(innat.logform) 31
Table7ResultsofModels 48

x
LISTOF FIGURES
Figure1DecreaseinRelativePrices(PPP) 16
Figure2TheEffect ofTerrorism ontheDemandforTourism 18
Figure3 IncreasesinGDPHandGDPSonthedemandfortourism 19
Figure4Histogramof TPC(innat.logform) 22
Figure5HistogramofTERR(innat.logfirm) 24
Figure6HistogramofPPP(innat.logform) 26
Figure7HistogramofGDPS(innat.logform) 28
Figure8HistogramofGDPH(innat.logform) 30

1
1.INTRODUCTION
A.ALookatTourism
Tourismhasbecome amajor economic activityinthe20thcentury.Thesupplyof
airplanesleftoverfromWorldWar IIledtothe growthoftheairline industry.Asthe
airlineindustrybecame morecompetitiveandtheworldeconomyexperiencedsubstantial
growth.Traveling,aluxuryonceenjoyedonlybythewealthy,becameaffordabletothe
masses.
Theeconomicsignificanceoftourismtodayisquiteevident.Tourismaccounts
for10%ofworldGDP.Table1summarizesthe growthinworldtourismsince1992.

2
Table1Tourism GrowthSince1950
YEAR InternationalTourist
Arrivals(000)
Tripsperthousand
worldpopulation
1950 25282 10.05
1960 69286 23.24
1965 112729 34.28
1970 159690 44.24
1975 214357 54.04
1980 287906 64.70
1985 329636 67.9
1990 455594 86.52
1991 455100 84.53
1992 475580 87.49
Tripshaveincreasednearlynine foldfrom1950-1992.Thisincrease reflectsthe
growingimportanceoftourismintheworldeconomy.
Thetourismindustrycanbenefitaneconomyinnumerousways.Tourism
increasesgovernmenttaxrevenue,reducesunemployment,andcreatesamorediversified
economy.Muchofthetourismindustryisconcentratedindevelopedcountries,andthe
majorityoftouristscomefromdevelopednations.However,tourismiswidely
consideredaviablesourceofincomefordevelopingcountries.Manyless developed
countriesare geographicallylocatedinareastoattracttourists.

3
CountriessuchasthoseintheCarribbeanhavetakenfulladvantageoftheirwarm
climateandbeachestoattractmillionsoftourists yearly.Tourismcomprisesupto50%
ofGDPinsomecountriesliketheBahamas.Thesecountrieshowever enjoy theunique
positionofbeingrelativelyclose geographicallytotheUnitedStates,alargedeveloped
country whosecitizenstravelextensively.Other developing countrieshavepotentialfor
economic growththroughtourismbutmaylacktheinfrastructureandresourcesfor
tourismdevelopment.
Limitedresourcesindevelopingcountriesconstrictthenumberofsectorsinthe
economy.Manydevelopingcountries exportmainlyagriculturalproducts. Investment
andlaborusedinagriculturalproductionmaycomeatthetourism?sindustry expense.
Thelackof resourcesmay forcedevelopingcountriestosubstitutebetweeneconomic
activities.
Easton(1998)studiedtherelationshipbetweentourismandcommoditytradein
Canadaandfoundevidenceofsubstitutabilitybetweentradeandtourismcitinglaborin
oneindustrytakesawaylaborfrom theotherindustry.Evidenceofsubstitutability
betweentradeandtourisminacountry asdevelopedasCanadasuggeststhatforless
developedcountriessubstitutabilitymaybe evenmorepronounced.
Manysmalldevelopingcountriesrelyonagriculturalexports,butheavy
relianceonagriculturalexportscanbevolatile.Droughtsandfluctuating worldprices
candamagean agriculturalbasedeconomy.Investmentinthetourismindustryenables
thedevelopingcountries todiversifytheireconomicbasebyrelyinglessonagriculture

4
andtoexploreotheropportunitiesforeconomic growth.Aspreviouslymentioned,
tourismcanincreaseGDP, foreigninvestment,andcreatejobs.
Yetlikeagriculture,tourismcanbesusceptibletooutsideinfluences.Natural
disasters,politicalinstability,terrorism,exchange ratefluctuations,domesticprices,and
pricevolatilitycanallinfluencethelocationfora vacation.Themainfocusof thisthesis
istofindwhat,ifany,effectterrorismhasontourism.Althoughseveralpapershave
previouslystudiedthisrelationship,I gofurtherby lookingatthedifferentialeffect
terrorismhasontourismbetweendevelopedanddeveloping countries.Thefollowing
sectiongives abriefoverviewofterrorismbefore itseconomiceffectonthetourism
industryisdiscussed.
B.ALookatTerrorism
Terrorismisthe?premeditateduseorthreatofuseofviolencebyindividualsor
subnationalgroupstoobtainapoliticalorsocialobjectivethroughtheintimidationofa
largeaudience,beyondthatoftheimmediatevictim.?(Enders,2003).
Byusingviolence againstnon-combatants,terroristsinstillfearincitizens.
Throughintimidation,theterrorist grouphopestoweakenthe government?slegitimacy
whilestrengtheningtheirownimportance.Theterroristgroupaccomplishesitsgoal(s)if
throughthethreator actionofviolencethe governmentsuccumbstotheirdemands,
usuallypolitical.
Defining whatconstitutesaterroristactisnecessaryinstudiesinvolvingterrorism
becausesourcesdefinethisnotiondifferently.Oftencomparingstatisticsonterrorism

5
fromtwodatasources revealverydifferentnumbersforidenticalevents.Thisthesis
adoptsthedefinitionofterrorismemployedbytheMemorial InstituteforthePrevention
ofTerrorism(MIPT).TheMIPT characterizesa terrorist groupas a group?belongingto
anautonomousnon-stateorsubnationalrevolutionaryoranti-government movement.?
Suchmovementsuseviolenceorthreatenviolencetoachievepoliticalgoals.The
violenceisusedagainstcivilianstocreatefear.
Besidesdeathanddestructioncausedbyterrorism,terroristactscanhaveserious
effectsontheeconomy.Tourismisonesectorof theeconomythatterrorismcan
immediatelyaffect.The followingsectiondescribeshowterrorismaffectsthetourism
industry.
C.Terrorism?sEffectonTourism
Terrorism canunravela country?stourismindustry.Terrorismdirectly affects
touristdecisionmaking.Touristsmaysubstitutebetweenvacationspotsifthey feel
threatenedorunsafeina country.Fewer annualtouristsasaresultofterrorismwill
typicallyresultinlossesoftourismrevenue.Lossesintourismrevenuemay havea
largerimpactinthosecountriesin whichtourismconstitutesalargerpercentageofGDP.
Terrorismcanalsoaffecttheamountofforeigninvestmentthatflowsintothe
country,especiallyinvestmentintourism.Typicallyinlessdevelopedcountrieswhere
resourcesarelimited,foreigninvestmentfundstourismbusinesses.Thethreatof
terrorismdiscouragesforeigninvestmentaslendersinvesttheirmoney elsewhere.

6
Finally,thethreatofterrorismforces governmentstoinvestmoreinsecuritythan
theywouldnormally.Thesefundscouldhavebeeninvestedelsewherethatmaybemore
beneficialtothe economy ifitwerenotforterrorism.Forexample,laborthatcouldhave
beenusedtocultivatelandsoraidinconstructionmustbetrainedtobecomepolice
officers.Theopportunity costsofdealingwithtourismmaybelarge relativetoavailable
sources.
D.ConcludingRemarks
Tourismhasbecome animportantsourceof worldGDPsincethe1950?s.Tourism
generatesrevenue,increasingforeigninvestmentandemploymentopportunities.The
tourismindustry canbedisruptedbyterrorismandtheuseofviolencecreates a
dangerousenvironmentfortourists.Terrorismaffectstourismbydecreasingrevenues
throughfewertourists,discouragingforeigninvestment.
Thisthesisstudiesterrorism?sdifferentialimpactontourismbetweendeveloped
anddevelopingcountries.Theoretically,a developedcountry?stourismindustrymight
sufferlessthanadevelopingcountry?stourismindustryfroma givenincidentsimply
because adevelopedcountryhas amorediverseeconomythanitslessdeveloped
neighbors.
Security addedbyadevelopedcountrywillcompriserelativity fewerresources
fromothereconomicsectorsthaninalessdevelopedcountry.Divertingfewerresources
allowsdevelopedcountriestocontinueinvesting andmarketingtheirtourismindustryat
similarpre-terrorismlevels.

7
Noformalinvestigationsintothedifferenceofterrorism?seffectontourism
betweendevelopedanddeveloping countries havebeenundertaken.Thisaspectofthe
resultsisuniqueandofferspotentialpolicyimplicationsforlesseningthedetrimental
effectsofterrorismontourism.
E.StudyOverview
Thesectionsofthepaperareas follows.Section2presentsaliteraturereview,
Section3methodology,Section4results,andSection5concluding remarks.The
literaturereviewanalyzesatypicaldemandmodelfortourismwhichthispaperbuilds
upon.
Section2alsolooksatotherliteratureonthetopicofterrorismandtourism,and
discussesthedifference betweenpreviousworkandthisthesis.Methodology(Section3)
includesanexplanationof econometrictestsonthedataset.Thissectionrelates
economictheorytotheindependentvariables.TheresultsinSection4are discussedin
the contextofeconomictheoryandcomparedtorelatedresultsinthe literature.
ConcludingremarksinSection5willofferpolicy suggestionsbasedonthe results.

8
2.LITERATUREREVIEW
A.PreviousLiterature
Terrorism?seffectontourismhasgainedtheattentionofUSeconomistssincethe
eventsofSeptember11.Thefollowingsectiondiscussessomeofthemore prominent
literatureonthistopic,beginningwithadiscussionofapaperby Brakke(2004) who
presentsamodelof tourismdemand.Iaddaterrorismvariabletoconstructamore
completemodelofthe demandfor tourism.
Includedintheliterature reviewisadiscussionof previousarticlesthatstudythe
priceelasticityofdemandinthetourismindustry.Mypaperfocusesonterrorism?seffect
ontourismandtheinclusionofindependentvariablesPPP(purchasingpowerparityor
therealpriceoftourism) andGDPofthehosttouristcountryprovides anopportunityto
analyzetheprice elasticity oftourism.
Brakke(2004) looksatUStourismdemandto85countriesoveraperiodof16
yearsfrom1984-1999andusesafixedeffectsmodelonapooleddataset.Thefixed
effectsmodelutilizesdummyvariablesinapaneldatasettomeasurefactorsthatare
constantovertimebutvary across crosssectiongroups.Thenumberoftouristarrivalsis
thedependentvariableinthemodel.Independentvariablesareaprice competitiveindex,
incomepercapitainthe source country ofthetourist,andapoliticalinstabilityvariable.

9
Theprice competitivevariablelooksattherelativepricebetweenthehostcountry and
thesource countryThesourcecountryistheUSandUSpricesareusedinthefollowing
calculation.ThedefinitionofpurchasingpowerparityisP=eP*wherePisthepricein
thetouristhostcountry,P*isthepriceinthe US,andeistheexchange ratebetweenthe
twocountries.BydividingPbyeP*andmultiplyingthistermby100wefind
competitivepricesacrosscountriesrelativetothe US. Theoretically,atouristshouldbe
attractedtoa countrythatisrelatively cheaper ceterisparibus.
Thisvariabledoesnotexplicitly includetransport costs.Transportcostsarenot
includedasaseparatevariable becausethisvariableisnotsignificantinotherstudies.
Transportcostswillnotbeofinterestina crosssectionalstudysincedistancesandfuel
pricesmaybesimilar acrosscountries.Further,airlineticketsarepurchasedinthe
countryoftouristoriginsounlessthisdatacanbe clearlyisolateditwouldnotbe
conclusive.
Demandfortourismwillriseasincomesofthesourcecountriesrise.Asincomes
rise,travelbecomesrelativelycheaperholdingpricesconstant.Potentialtouristswill
havemoredisposableincometospendonvacations.
Thepoliticalinstabilityvariableof Braake comes from Freedom HouseCountry
Ratings (2005)Countriesarerated1to7onpoliticalrightsandcivilliberties.Alower
numberindicates more freedom.Politicalstrifecandetertourismby creatingan
unpleasantandunsafeatmosphere.
Braake foundthatincomepercapitaofthesourcecountry hasapositiveeffecton
tourism demand,andpoliticalinstabilityhasanegativeeffectontourismdemandas

10
theorysuggests.Relativeprices,ontheotherhand,arenotsignificant.Braake explains
thisresultisareflectionofconsumertastes,andthisconclusionisreflectedbythe
dominanceofhighlydevelopednationstoattracttourists.Developednationsare
relativelymoreexpensivethandevelopingtourist nationsbutmustbemoreableto
accommodatetouristpreferences.
Thedemandmodelusedinthisthesisisbaseduponthemodelcomposedby
Braakewitha few exceptions.Aspreviouslymentioned,aterrorismvariablereplacesthe
politicalinstabilityvariableusedbyBraake.Problemswithmulticollinearitywillariseif
bothareincludedinmy modelbecauseterrorismaffectsthepoliticalstabilityofanarea.
Also,GDPpercapitaofthehostcountry(touristdestination) isanindependentvariable
inthismodel.Theinclusionofthisvariableisunique.Braake?sstudyandotherstudies
concentrateontheGDP ofthesource countries toexplaintouristdemand.GDPper
capitaofthehostcountryisaproxy forthelevelofcomfortandqualityoftourist
amenitiesthatacountry has.Obviously,the abilityofa countrytoprovidefortheneeds
oftouristswillincreasedemandfortourisminthatcountry.
Thenegative effectofterrorismontourismispersistentthroughouttheliterature.
Doterrorismincidencesimpacttourismpermanentlyorhaveshortterm effects?This
questionisexploredby AllyandStrazichick(2000).ThesetwoeconomistslookatEgypt
andIsrael,twocountries sufferingfromterrorismthatnevertheless continuetoattracta
largenumberoftourists.TheannualdatafromEgyptcoversthe years1955-1997.This
datawascollectedfromtheBankofEgypt?sAnnual EconomicBulletin.Annual Israel
datacoversthe year1971-1997andcomes fromthe IsraelMinistryofTourism.Usinga

11
two-breakminimum Langrangemultiplier unit-roottestdesignedbyStrazichik,theytest
whetherthedataisstationary.Astationarymeanreverting resultsuggests that
terrorism?seffectontourismistransitory whilea non-stationaryresultshowsthattourism
neverfullyrecoversfromaterroristact.Theirtestsprovidedstationaryresultsindicating
theeffectsofterrorismarenotpermanent.
DrakosandKutan(2003)studytheeffectsofterrorismontourisminGreece,
Turkey,andIsraelusing monthlydatafromJanuary1996toDecember1999.These
countriesreceivelargenumbersoftouristsdespitebeingsusceptibletoterroristattacks.
Usingaseeminglyunrelatedregressionmodel,DrakosandKutanfindthat terrorism
negativelyaffectstourismandsubstitutioneffectsexistbetweenthesecountries.A
highernumberofterroristattacksinIsraelresultsinanincreaseintouriststoGreece.
DrakosandKutanalsofindthatintensityofterroristattacksaffecttouristdecisions.
EndersandSandler(1991)studythe effectthattheETAterroristgrouphashad
onSpanishtourism.Spainisconsistentlyinthetopfivevisitedcountriesintheworld.
Between1985and1987,ETAspecificallytargetedtouriststodamagetheindustry.The
authorsemploy avector autoregressivemodelonmonthlydataduring1970-1991.
EndersandSandersfindthateachtransnationalterroristattackdissuaded140,000tourists
fromvisitingSpain,resultingindecreasedrevenueswhenmultipliedbyaveragetourist
expenditure(avaluenot calculatedinthepaper.)
Enders,Sandler,andParise(1992)use anARIMAmodelwithatransferfunction
tostudylaggedeffectsofterrorismontourismfor Austria,Greece,andItaly. Lookingat
tourism receiptsduring1974-1988,theauthorsfindsignificantnegativelaggedeffectsof

12
terrorismonthetouristindustry.Theauthorsfindlossesoftourismrevenuefor
continentalEurope amountingto16billionSDR?s.
Thefollowingliterature reviewdiscussesthepriceelasticityofthetourism
industry.Garin-MunozandAmaral(2000)measurethedemandforSpanishtourism
servicesusinganunbalancedpaneldatasetof17countriesfrom1985-1995.The17
countriesarethetouristsourcecountries.MostofthecountriesareEuropean,however
theUS,Canada,andMexicoarealsoincludedinthedataset.Theunbalanceddataset
allowsfordifferentnumberofobservationsforthesource countries.The authorsemploy
OLSonasetof yearlydata.
Thedependentvariableusedinthepaperisthenumberofnightspercapitaspent
inhotelsbytouristcountryoforigin.Independentvariablesincluderealpercapita
income,exchangerates,andrealprices.Realper capitaincomeofthetouristsource
countrieswerecalculatedbyexpressingGNPofeachcountryin1990U.S.dollarsand
dividingbythepopulationofeachcountry.Transportationcostswere excludedfromthe
modelduetolackofdata.
Theresultsindicatethat allrealincomepercapitaandexchangeratesare
significant atthe10%level.Themodelwasindoublelogformsothatresultscouldbe
expressedinelasticities.Realincomeper capitaandexchangerates arepositively
correlatedwithnightsspentinhotelswithcoefficientestimatesof1.4and0.5.These
resultsare across allcountries.

13
Thepositiveresultsofthesevariablessuggestthatasincomesriseinthesource
countries,peopletendtotravelmoretoSpain,andthatadpereciationoftheSpanish
currency(atthattimethepeseta)increasesthedemandfortourismaswouldbeexpected.
B.Conclusion
Thepreviousliteratureusestimeseriesorpaneldataandfocusesonindividual
countryor regionaleffectsofterrorismontourism.Terrorism?snegativeimpacton
tourismistheprevalentconclusioninpreviousstudies.Topics includingthelongevity
ofterrorism?simpactontourismhavebeenstudied.Thepreviousliterature hasnot,
however,lookedatdifferencesbetweendevelopedanddevelopingnations,thesubjectof
thisthesis.
Also,thepreviousliteratureconcludesthattourismisapriceinelasticindustry.
Theresultsofthisthesisyieldsimilarresults.Thepreviousliterature focuseson
individualcountrystudies,but Ifindthatpriceinelasticityoftourismholdsacrossa wide
rangeof countriesfromvariousregions.The resultsaremore generalandpolicy
suggestionsperhapsmorerelevantthanwouldbe forcountriesnotindividuallystudied.

14
3.METHODOLOGY
Unlikepreviousresearch,thisthesisisacrosssectionalstudywhileotherpapers
studyingtheeffectsofterrorismusetimeseriesdata.Byusingcrosssectiontechniques I
canmake general,cross country commentsonterrorism.Theeffectsofterrorismcan
havedevastating effects ontourismbutdevelopednationsmaysuffer fewernegative
effectsthanpoorernationsduetoeconomic diversity.
Crosssectionaltechniquesareutilizedtodistinguishthisdifference.Thisthesisis
acrosssectionalstudyof60countriesduring2003.Allvariablesinthefinalregression
areinlogformtoreporttourismdemandelasticitiesofindependentvariables.
A.Model
Thefundamentalregressionequationexaminedis:
TPC=a0+a1PPP+a2TER+a3GDPS+a4GDPH+a4INT+ e
whereTPCistouristarrivalspercapita,PPPispurchasingpowerparity,TERisthe
numberofterroristattacks,GDPSisgrossdomesticproduct percapitaofthetourist
sourcecountry,GDPHisgrossdomesticproduct percapitaofthehostcountry,andINT
isaninteractivetermcombiningGDPHandTERR.Theerrorterm,e,isassumedtobe
normallydistributed.

15
Thebasisforthismodel comesfrom Braake(2004).Purchasingpowerparity,
PPP,andGDPSaretheindependentvariablesfromBraake?stourismdemandmodel.
Braake alsoincludesapoliticalinstabilityvariable.Theinclusionofapoliticalinstability
variableandterrorismvariablemayleadtoproblemsofmulticollinearityandpolitical
instabilityis notincludedinthepresentstudy.
Thedependentvariable TPCistouristarrivalspercapitafor2003.Thisvariable
dividesthetotalnumberoftouristsvisitingeachcountryinthesamplebythepopulation
ofthecountry.The arrivalspercapitavariable standardizesthevariable acrosscountries.
Somecountrieslikethe USattractahighnumberoftouristsbutrelylittleontourismdue
tosizeand diversity.TPCintheUSissmall relativetocountrieslikethe Bahamasthat
attractfewertouristsbut arrivals relative tothepopulationintheBahamas islarge.
Theassumptionisthattourismismoreimportanttocountriesthatattractlarge
numbersoftouristsrelativetothepopulation.Thesources forthedatainclude Tourism
MarketTrends(2004) forthenumberoftouristsandthe PennWorldTablesforthe
populationforsamplecountries.
TheindependentvariablePPP,oneofthevariablesinBraake (2004),isthereal
exchangerateofthehostcountry.Thisvariableislistedinthe PennWorldTables.
Purchasingpowerparity isdefinedasPPP=P/eP*wherePis theUSpriceofthetourist
sourcecountry,P*thepriceof thehostcountry,andeistheexchangerate expressedas
$/currency.PPPtakesintoaccountpricesinbothcountriesaswellastheexchangerate
betweenthe currenciesinthecountries.The USisusedasthebasecountry to
standardizerelativepricesacrosscountries.ThePPPvaluefortheUSis1.Acountry

16
withaPPPvaluesmallerthan1isrelativelymore expensivethanthe US,andacountry
withaPPPvaluelargerthan1wouldberelativelycheapcomparedtothe US.
Economictheorysuggeststouristswillbeattractedtocountrieswithlowerrelative
prices.Lowpriceswillincreasethequantityof tourismdemandedandattractvisitors.
Thisincreaseinthequantitydemandedfor tourismisdepictedinFigure2.P1represents
initialpricescorrespondingwithQ1,withthenumberoftouristspurchasingtourismatP1.
ThedecreaseintherelativepricetoP2correspondswiththeincreaseinquantity
demandedfortourismatQ2.
Figure2.DecreaseinRelative Prices(PPP)
Price
1
P2
Q1Q2 Quantity
One assumesthisprincipletranslatesonaverage intotouristsfromdeveloped
nationstravelingtodevelopingnationswherepricesare cheaper.Poor countrieslike
ThailandandMoroccoattractmanyEuropeantouristsfromdevelopednationssuchas
FranceandEngland.Myresultswillshowthattourismisdominatedbydeveloped
nations.Themajorityof touristscomefromdevelopednationsandtraveltodeveloped
nations.Relativeprices maybemore pertinenttotimeseriesstudieswherechangesin
pricescanbemeasuredover timeandtouriststoa particular countrymight respond.

17
ThevariableGDPSistheGDPpercapitaofthesourcecountry.Thesource
countryinthepresentthesisisthecountryoforiginofthelargestnumbersoftourists.As
Braake(2004)pointsout,higherincomesdecreasethe shareoftravel expenserelativeto
incomeallowinglongerandmorefrequentopportunitiestotravel.Assumingtourismisa
normalgood,apositivecoefficientonGDPS willreflectanincreaseofdemandfor
tourismasincomesinthesourcecountry grow.
Sourcecountries compiledfromthe Compendium ofTouristStatistics(1996)
includethetopsource countryoftouristsforeachhostcountry forthe year 1996.The
base yearofthisthesisis2003.Duetolackofdata,source countrydatais from1996.
GDPofthesourcecountries,fromthe PennWorldTables,areexpressedin2003US
dollars.
Thelasttwoindependentvariables,terrorismandGDPofthehostcountry,are
variablesthepresentthesisaddstothe Braaketourismdemandmodel.Thesevariables
arealsothe focusofthis thesis.
Terrorisminstillsfearincitizensandvisitors,causesdamage,anddiverts
resourcestoaddedsecurity.Obviously,terrorismshouldnegativelyimpactmanysectors
oftheeconomyincludingtourism.Terroristsoftendeliberately attacktouriststoreceive
greatermediaattention.TheBalinightclubbombingkilledmanytouristsanddamaged
Bali?stourismindustry (Henderson2003).Touristscanbeexpectedtosubstitute
vacationdestinationsorjuststayathomeifthey feelthreatenedincertaincountries.
Terrorismshoulddecreasethedemandfortourismata givenprice(P)astouristswill
substitutevacationdestinations orsimplystayhome,shiftingthedemandcurvefromD0

18
toD1inFigure3.Thesupplycurve(S)isassumedtobeperfectlyelastic meaningthatat
agivenpriceoftourism,countriesarewillingandabletoaccommodateanynumberof
tourists.
Figure3.TheEffectofTerrorismontheDemandforTourism
Price
D1D0
P S
Quantity
Theterrorismvariableincludesthetotalnumberofterrorismincidences(domestic
andinternational)thatoccurredineachcountry duringthe five yearspanfrom1999-
2003.Five yearsofterrorismincidencesareincludedinthedatasinceviolentattacks
occurringinpast yearsmayhave aneffectoncurrenttouristdecisionmaking.
Theothervariableofinterest isincomeper capita ofthehostcountry,GDPH.The
previousliteratureshowsthattourismisasectordominatedbydevelopedcountries.
Developedhostcountriesmaybemoreabletoaccommodatetouriststhanpoorernations.
Thisvariablerepresentsareflectionofcomfortandperhapssafety.Greatercomfortand

19
touristamenitieswillincreasetouristdemandforthehostcountryresulting inmore
visitors.
GDPofthehostcountry (qualityoftourismservices)wasnotincludedinthe
studybyBraakewhoinsteadincludesGDPofthe touristsourcecountryarguingthat
tourismisanormalgood.Asincomesrise,more moneywillbespentontourism.I
includebothGDPofthe sourceandhost countriesforreasonsmentionedabove.The
resultsindicateeconomicsignificanceof GDPH validatingitsinclusioninthemodel.
IncreasesinGDPs andGDPHshouldincreasethe demandfortourismata givenprice,
indicatedbytheshiftfromD0toD1 inthefollowingfigure.
Figure1.Increasesin GDPHandGDPSonthedemandfortourism
Priceof D0 1
Tourism
Quantity
TheabilityforhighGDPHcountriestoprovidefortouristdemandsmaybe
reflectedinhigherrelativeprices(PPP)indevelopedcountries.However,iftouristsare
concernedwithprices,thenonewillexpectPPPtobepositiveandstatisticallysignificant

20
signifyingtheability for lessexpensivecountriestoattractalarger groupoftouriststhan
expensivecountries.Thisresultwouldrevealthattouristsarepricesensitive.
Developingcountries areusually cheaperthandevelopedones,butexpensive
developednationsare betterabletoaccommodate tourists.Theoretically,developedand
developingnationswillattracttouristsforopposingreasons,anditisunlikelythatboth
variables(GDPHandPPP)willbesignificant.Theresultswillrevealtouristpreferences.
Thelastvariable INTisaninteractivevariablebetweenterrorismandGDP ofthe
hostcountry (TERR*GDPH). In additiontotheeffectofterrorismontourism,a
contributionofthis thesisistolookatthedifferenceintheeffectofterrorismonthe
economiesofdevelopedanddevelopingnations.Theoretically,terrorism willharma
poornationmorethana developedcountry.Adevelopedcountryhas amorediverse
economyandtourismisrelativelylessimportanttotheeconomythanina poorernation.
Definitely,theeffectsof 9/11hadnegativeeffectsonUS tourismandeconomy(Looney
2002)buttheUSwasperhaps abletoreboundeconomicallymuchquicker thanifthis
attackhadoccurredinasmaller,lessdevelopedcountry.Apositiveinteractive
coefficientwillindicatethathigherGDPofthehostcountrywilloffset,atleast
somewhat,thenegative effectsof aterroristattack.
B.Data
ThedataonterroristincidenceswascollectedfromtheMemorial Instituteforthe
PreventionofTerrorism(MIPT).Thedataincludesbothdomesticandinternational
incidences.Itmaybearguedthatadomesticterroristincidenttargetingthegovernment

21
maynotdiscourage atouristfromvisitingthecountrysincethetouristisnotthetarget
butdomesticterrorismisasignalofpoliticalinstability andcancreateanunsafe
environmentforavisitor.Thecollectionofthisdatainvolvedphysically countingthe
numberofterrorismincidencesforeachcountry.
Tourismsourcedata was collectedfromthe CompendiumofTourist Statistics
(2003). Touristarrivaldatacomesfrom TourismMarketTrends(2004).Thedatainthe
publicationswasgatheredbytheWorldTradeOrganization(WTO).Dataontourism
maybe unreliable anditisnecessarythatalldata comefromsimilarsources.Countries
may exaggeratetourismrevenueformarketing reasons.TheWTOiswidely recognized
asaviablesourcefortouristinformation.TheWTOdefinesatouristas?apersonwho
travelstoanotherplace,outsidetheirusualenvironment,forprivateinterestsorissent
there(bywork)butisnotemployedatthisplace.? (2006)
C.DescriptiveStatistics
Descriptivestatisticsofalldependentandindependentvariables areuseful
backgroundinformationandmayindicatefutureproblemsincoefficientestimates.
Ordinary LeastSquares (OLS) estimationisusedinthispaper.Oneofthe assumptions
ofOLSisthatdependentvariablesarenormallydistributed.Non-normaldistribution
maybeduetooutliersandsoon.Modelmisspecificationisaproblemresultingfrom
non-normaldistributedvariables.
TheRamseyReset Testexaminesmisspecificationofthemodel.Amisspecified
modelwillproducebiasedandinconsistentestimatesthusthefollowinganalysisprovides

22
asummaryofdescriptivestatisticsandhistogramsforallvariablestoprovidereasonfor
futureeconometrictests.
Thedependent variable,touristarrivalsper capita,indicatesthedemandfor
tourisminthehostcountry.Thefollowinghistogramrevealsthedistributionofthe
dependentvariableforthe60countries.The appendixattheendofthepaperincludes
theentiredataset. Thedescriptivestatisticsfollowthehistogramsothatinterpretationof
thedistributionispossible.Thehistogramsanddescriptivestatisticsofthevariablesall
reportthenaturallog formofthevariablessincetheregressionisinlogform.Figure4is
ahistogramoftouristarrivalsandTable2liststhe descriptivestatisticsfortourism
arrivalspercapita.
Figure4.HistogramofTPC(innaturallogform)
Number
Of
Countries
LNARRPCAP

23
Table2DescriptiveStatisticsforTPC (innat.logform)
MEAN -2.3
STANDARD
ERROR
.23
MEDIAN -2.14
STANDARD
DEVIATION
-1.8
SAMPLE
VARIANCE
3.2
KURTOSIS -.7
SKEWNESS -.2
RANGE -7.4
MINIMUM -6.3
MAXIMUM 1.1
COUNT 60
Themeanof -2.3correspondswiththevalueof0.005.Theaveragetourist
arrivalspercapitaisnot verylargeandperhapsforthemajorityofcountriesinthe
sampletourismdoesnotcontributegreatlytothe economy.
Thehistogramreveals,asthedescriptivestatisticsindicate,aslightnegativeskew
andnegativekurtosis.Skewnessmeasuresthesymmetryofthedata.Anormally

24
distributeddatasetwillpeakatthemeanwherethemajorityoftheobservationsliethen
decayrapidlyontherightandleftsides.Theobservationsaretailedleftwardsexplaining
thenegativeskew.Kurtosis,measuringthepeakofthedataset atthemeanislowfor
thissetofdata.Asthedescriptivestatisticsshow,thekurtosisis-0.7.Thereisnopeak
inthisdata,ratherthedataappearstocavein.Despitethelowkurtosis,thestandard
deviationisrelativelylowthereforeasignificant numberofobservationsdoliecloseto
themeanwhichiswhat weshouldexpect.
Nextconsidertheindependentvariableofprimaryinterestinthestudy,terrorism
incidences.Figure5isa histogramofterrorismincidences anddescriptivestatisticsare
showninTable3.
Figure 5.HistogramofTERR(innat.logform)
Number
of
Countries
LNTERR

25
Table3DescriptiveStatisticsforTERR(innat.logform)
Theaveragenumberofterrorismincidencesacrossallcountriesis120.9.Thisis
anaverageofabout24attacksper yearpercountry overthefive yearperiodcoveredby
thedata.Theskewnessofthedataislow,0.23,indicatedbythesymmetry oftheright
andleftsidesofthehistogram.Unfortunatelythe tailsofthedistributionareveryfat,
whichexplainsthenegativevalueofthekurtosis.Thestandarddeviationisquitelarge
COUNT 60
STANDARDERROR .28
MEDIAN 2.5
MODE 0
STANDARDDEVIATION 2.13
SAMPLEVARIANCE 4.54
KURTOSIS -1.08
SKEWNESS .23
RANGE 7.03
MINIMUM 0
MAXIMUM 7.03
SUM 177.3
COUNT 60

26
relativetothemeanindicatingalargedispersioninterrorismincidencesamongthe
countriesinthesample.CountriessufferingmostfromterrorismincludeColombiaand
Spain,with1135and732terroristincidencesrespectivelyacrossthis5yearperiod.
AustraliaandPortugalareexamplesoftwocountriesinthesamplenotaffected
byterrorismwithoneattackeach.
ThenextvariabletoencounterisPPP(purchasing powerparity).Figure6isa
visualrepresentationofPPPanddescriptivestatisticsofPPParelistedinTable4.This
dataprovidesadequatebackgroundinformationonthisvariablepertaining tothe
countriesinthedataset.
Figure6.HistogramofPPP (innat.logform)
Number
of
Countries
LNPPP

27
Table4DescriptiveStatisticsof PPP(innat.logform)
MEAN 2.55
STANDARDERROR 0.39
MEDIAN 2.03
MODE 2.48
STANDARDDEVIATION 3.0
SAMPLEVARIANCE 8.95
KURTOSIS 1.76
SKEWNESS 1.22
RANGE 14.86
MINIMUM -1.2
MAXIMUM 13.66
SUM 153.26
COUNT 60
Thepositivemeanvalue indicatesthemajorityof thecountriesinthesampleare
cheaperthantheUSasPPParerelativepricesbetweenthehostcountriesandtheUS.
TheUSisgiventhevalueof1.Countrieswhere PPP<1aremoreexpensivethanthe
USwhilethosecountrieswherePPP>1arerelatively cheaper.Thenaturallogofoneis
zerothusthepositivemeanvalue indicates relatively cheaperpricesthantheUSon
averageinthesample countries.

28
Thepositivekurtosisindicatesa ?peak?inthedatawithrapiddecayoneitherside
ofthepeakhoweverthe dataishighlyskewedtotheright asindicatedbytheskewness
value1.22.Aquicklookatthehistogramwillpointouttherighttail.The dataisspread
outoversuchalargedistancethatthestandarddeviationislargerthanthe mean.
ThenextvariableisGDPs(GDPofthesourcecountry).Thehistogramis
representedinFigure7anddescriptivestatisticsforthisvariablearelistedinTable5.
Figure7.HistogramofGDPS(innat.logform)
Number
of
Countries
LNGDPSOURCE

29
Table5DescriptiveStatisticsofGDPS (innat.logform)
MEAN 9.94
STANDARDERROR .09
MEDIAN 10.21
MODE 10.21
STANDARDDEVIATION .67
SAMPLEVARIANCE .45
KURTOSIS 2.23
SKEWNESS -1.71
RANGE 2.75
MINIMUM 7.78
MAXIMUM 10.53
SUM 596.45
COUNT 60
Theincomeofthesourcecountriesisonaverageveryhigh. Itisworthnoting
manyofthesourcecountries suchasGermany andUSappearseveraltimesinthedata.
TheUSandGermanyarerichcountrieswithlargepopulationswhosecitizensenjoy
traveling,accounting for theirnumerousappearancesintheGDPSdataset.Themultiple
observationsofGermany,the US,andotherdevelopedcountriesaccountsforthelarge
positivekurtosisor?peak?inthehistogramaswellasthenegativeskew.Mostofthe
sourcecountriesarelocatednearthehighmeanlevelwithpoorersourcecountries

30
makingupthenegativeskew.Ascanbeexpectedwithsomanyobservationsaroundthe
meanthestandarddeviationisverylow.
Therangeofhostcountry income GDPHobservationsvary widelyacross
countriesassomearehighlydevelopedandothersarepoordevelopingcountries.The
histogramshowingGDPHobservationsisdepictedinFigure8anddescriptivestatistics
areinTable6.
Figure8.HistogramofGDPH(innat.logform)
Number
of
Countries
LNGDPH

31
Table6DescriptiveStatisticsofGDPH(innat.logform)
MEAN 8.98
STANDARDERROR .13
MEDIAN 8.95
MODE N/A
STANDARDDEVIATION .98
SAMPLEVARIANCE .97
KURTOSIS -.32
SKEWNESS -.42
RANGE 4.10
MINIMUM 6.42
MAXIMUM 10.53
SUM 538.85
COUNT 60
Themeanof GDPforthehostcountriesisquitehighwithalowstandard
deviationasthehistogramindicatesalargenumberofobservationsaroundthemean.
Thedataishoweverskewedtotheleftasthenegativeskewimplies.The lefttailinthe
datasetrepresentslowincomecountries.Manyofthelowincomecountriesinthedata
setarelocatedinAfrica.Thekurtosisislowindicatingthelackofapeak.Clearlythe
righttailofthedataincludesmoreobservationsthanaroundthemeancausinganon-
normaldistribution.

32
D.Jarque-BeraNormalityTest
TheJarque-BeraTestteststhenormalityassumption,takingintoaccountthe
skewnessandkurtosisof theresidualsoftheregression.Thenullhypothesisisthatthe
residualsarenormallydistributed.TheJarque-Berafollowsachi-squareddistribution
withtwodegreesoffreedom.The5%criticalvalueis5.99.TheJarque-Berastatisticis
calculatedinthe following equation,
(T/6)S2+(T/6)(((k-3)2)/4)
whereSisskewness,kiskurtosis,andTisthenumberofobservations.Failuretoreject
thehypothesis(Jarque-Beravaluelessthan5.99) meansthattheresiduals ofthe
regressionarenormally distributed.Thevaluesofkurtosisandskewnessoftheresiduals
fromthetourismmodelare-.29and-.47respectively.ThecalculatedJBstatisticis
29.27,thusthe nullhypothesisisrejectedandtheresidualsarenotnormallydistributed.
Thisresultisnotasurprisegiventheskewedvariables. Theresidualswouldprobably
resembleamorenormal distributionwithalarger dataset.Thisisashortcomingofthe
data,howeverthemodel passes misspecificationtests.

33
4.RESULTS
A.Introduction
Theresultsofthispaper willprovideinsightintotheterrorismand tourism
industry.AlleconometricanalysiswasdoneusingMicrosoftExcel2007andLimdep.
The Limdepresultsareincludedintheappendix.Variables areinnaturallogssothat
elasticitiescanbe reported.
B.TablesandResults
Table4.A,onpage48,providescoefficientestimatesandt-statistics.The R2 of
theregression is0.57.Themodelexplains57%ofthevariationintourismdemand
acrosscountries.Thisseemsreasonable giventhearrayof countries representedinthe
sample.
Therealexchangerate(PPP)isnotsignificant suggestingtouristsareinsensitive
torelativeprices.Thisiscontrarytoeconomicintuitionthathigherprices willdeter
tourism.However countrieslikeSpain,France,and Italyhaveremainedamongthetop
touristdestinationsevery yeardespite becomingrelatively expensiveespeciallywiththe
recent euroappreciation.Thesecountries are richinhistoryandhavemanytourist
attractions.Theaverage touristisevidentlymore concernedwith

34
anenjoyabletimethanthecheapestdeal.Thesetoptouristdestinationsareableto
accommodatethe demandsoftouristssothetourismindustrymaynotbeassensitiveto
pricesassomeotherindustries.Thisresultindicatesthattouristsmaynotsubstitutea
cheapervacationfor amoreexpensivevacation.Amodelusingtourismrevenuesasthe
dependentvariableisdevelopedlatertocomment specificallyontheprice elasticityof
thetourismindustry.
Terrorismissignificantatthe5%levelwithat-statisticof -2.16. Thecriticalt-
statisticis -1.67. Thecoefficient estimateofterrorismis -1.67.A1%increaseinthe
numberofterrorismattacksdecreasesthedependentvariabletourismarrivalspercapita
(TPC)by1.67%.
Theeffectofterrorismontourismisverylarge eventhoughthe chanceof being
thevictimofaterroristattackissmalleveninacountryexperiencinga highnumberof
terroristattacks.Theability toinstillfearincitizensandvisitorsisoneofthegoalsof
terrorist groups.Bydamagingthetourismsector,governmentsaremorelikelyto
concedetoterroristdemandsespeciallyincountrieswheretourismisvitaltothe
economy.
Touristsaresensitivetoterroristattacksandconsidersafety aprioritywhen
choosingwheretotravel. Itmustbenotedthatthiscoefficientestimatedoesnotinclude
theinteractivevariable.I arguethatGDPofthehostcountrywillaffectterrorism?seffect
ontourism.Theoverall effectwillbediscussedwithinteractivevariable.
TheGDPSper capitaofthesource countryisnotsignificantwithat-statisticof
0.25.Themostlikelyreasonisthattouristsvisitinganareacome frommanycountries

35
whilethenumberoftouristsformtheprimarysourcecountriesmaycomprisefewerthan
10%oftotalvisitors.
Also,GDPS measuresthe incomeinthemajorsourcecountry.Mosttourists,
however,may earnaboveaverageincomes,especiallythosetouriststravellingalong
distance.GDPsmaynot accuratelyrepresenttheaverageincomesof actual tourists.
Incomeofthesource countryisstatisticallysignificantinBraake?stourism
demandmodel.EconomictheorysuggeststhatGDPHshouldbeincludedwithother
independentvariables.MostpreviousliteraturefocusesonGDPS.Theinclusionof
GDPHinthemodelisuniquetothepresentstudy.
Tourismisanindustrythatisdominatedbydevelopedcountriesassupportedby
the statisticalsignificanceof GDPH.Thevariable?st-statisticisstatistically significant at
the1%criticalvalue.The1%criticalvalueis2.40andt-statisticforGDPHis2.45.The
economiceffectofGDPHonTPCmayincludethe interactivevariablediscussedbelow.
There arenumerousreasonswhyGDPHwillaffecttourismdemand.As
previouslymentioned,wealthiercountries canprovidemorecomfort andsafety for
tourists withinvestmentinthetourism industry.Thesecountriesmaybeabletolure
moreforeigndirectinvestmentaspotentialinvestorswillbeattractedbyprofit.
Developedcountrieshavemoreresourcestospendonpromotingtheircountryas atourist
destination.Thisresultshowsthattourismisanindustrythatisdominatedbythe
developedworld.
ThestatisticalsignificanceofGDPH gives greaterinsightintotouristpreferences.
PPP?sstatisticalinsignificanceconcludesthattouristsarenotattractedtocheapvacation

36
destinations.GDPH?sstatisticalsignificanceshowsthattouristsareattractedtoareasthat
providegreatercomforttotouristsandarelikely moreexpensive.Insensitivitytopriceis
consistentwithattractiontodevelopedcountries.
Theinteractivevariable combinesterrorismandGDPH.Thet-scoreof1.96
indicatesstatisticalsignificance atthe5%level.Theelasticity coefficientis0.16.We
canconcludebythestatisticalsignificancethatGDPHdoesaffectterrorism?seffecton
tourism.Theleveloftheeffectisdiscussedbelow.
Terrorismcanseriously damagea country?simageasavacationdestination.
Tourismisrelativelylessimportanttodevelopedcountriesthanpoornationsduetotheir
economicdiversity.Developingcountriessufferingfromterroristattacks mustdivert
resourcesformothereconomicsectorstoinvestinsecurity.Thisisanecessarymeasure
yetthediversionofresourcescantake atollonafragileeconomy.Developedcountries,
ontheotherhand,canincreasesecurity withoutmuchaffectingothereconomicsectors.
Awidersourceof fundsenablesdevelopednationstorepairtheirimageas atourist
destinationmuchmoresuccessfullythanlessdevelopedcountry.
Theinteractivetermmustbeincludedwhenderivingthefulleffectofterrorism
andGDPofthehostcountryontouristarrivalspercapita.Thecoefficient ofterrorismis
-1.66yet wemustincludehowGDPofthehostcountryaffectstheeffectofterrorismon
arrivalspercapita.Toderivethetotaleffectofterrorismontourism,solvethefollowing
partialderivative:
?TPC/dTERR=-1.66+0.16MGDPH

37
whereMGDPH isthemeanoftheGDPpercapita ofthehostcountriesacrossall
countries.Themeanof GDPHiscalculatedbydividingthesumofallGDPHbythe
numberofcountriesinthedataset.Thevalueofthemeanis8.98.Taking intoaccount
GDPH?seffectofterrorismontourism ?TPC/?TERR=-0.22.A1%increasein
terrorismattacksdecreasestouristarrivalsper capitaby0.22%.
Towhatextentdoestheeffectofterrorismontourismdifferbetweendeveloped
anddevelopingcountries?Byusingthestandarddeviationthisdifference canbe
calculated.ThemeanvalueofGDPHis8.98withastandarddeviationof0.98.Addand
subtractthestandarddeviationfromthemeanvalue toseparatedevelopedfrom
developing countries.Solve ?APC/?TERtofindtheeffectonterrorismontourismfor
whatcanbecalleddevelopedanddeveloping countries.Onestandarddeviation
comprises67%ofthedata.
Theresultsfromthiscalculationindicatecoefficientsof -0.07fordeveloped
countriesand-0.38fordeveloping countries.Fordevelopedcountries,a1%increasein
terrorismincidencesdecreasestouristarrivalsper capitaby0.07%whiledeveloping
countriessuffera0.38% decreaseintouristarrivalsduetoa1%increaseinterrorist
attacks.Thedifferenceissubstantial.Theeffect ofa1%increaseinterrorismontourist
arrivalsisnearly6times greaterindevelopingcountries.
Iftwostandarddeviationswereusedinthecalculationover90%ofthe countries
inthesamplewouldbeincluded.Followingthislogic,itisquitelikelythattheverytop
touristcountriesarenotaffectedatallbyterrorismwhilethebottomcountriesare grossly
affected.

38
Likewise,tofindthetotaleffectofGDPHontourismarrivals,wemustalso
include terrorism?s effectontheeffectthatGDPH hasontourism.Touristsmaybelieve
moredevelopedcountriesaresafer andmore comfortable,increasingthedemandfor
tourisminthatcountry.Terrorismcandamagethatimage.ThefulleffectofGDPHon
touristarrivalsper capita thefollowingpartialderivativemustbesolved:
?APC/?GDPH=.7+.16MTERR
whereMTERRisthemeanofterrorismattacksacrosscountriesinthesample.The
derivedcoefficientis1.17.A1%increaseinGDPHincreases arrivalsper capitaby
1.17%.Aterrorismattackincreasestherelativeimportanceofsafetytotourists.GDPH
isusedasaproxyforthequalityandquantityoftourismservices,andthe safetyfactoris
amajorcomponentoftourismservices.Therefore,aterroristattackonacountry
increasestheeffectofGDPH ontourismarrivals.Thisresultisafurtherindicationof
consumer?spreferenceofqualityvacationoverbargainvacation.
C.MoreEvidenceofDifferencesbetweenDevelopedandDevelopingCountries
TheChowtestprovidesfurtherevidenceofdifferencesbetweendevelopedand
developing countrieswithregardstoterrorism?seffectontourism.Asmentioned
previously,roughlyhalfthesamplecountries are considereddevelopedwhiletheothers
aredeveloping.AlthoughtheChowTestcannotreportelasticities,theresultscanbe
usedtosupporttheresultsobtainedfromtheOLS analysis.TheChowTestcomparesthe
errorsumofsquares fromthetwosubsamplestoseeifthetwogroups are statistically
different (havedifferent slopesandintercepts).

39
Oneofthesubsamplesisconstrainedwhiletheotherisunconstrained.The
constrainedsubsample assumesthereisn?tadifferencebetweendevelopedand
developing countries?tourismeconomies.Therefore,thereisnointeractionordummy
termintheconstrainedmodelsinceGDPHdoesnotaffectterrorism?s effect ontourism.
Developedanddevelopingcountriesareconstrainedbecausetheyareforcedtohavethe
sameslopeandintercept.Thefollowingisadepictionoftheconstrainedsubsample.
ATC=B1 +B2PPP+B3TERR+B4GDPS +GDPH
Theunconstrainedsubsampleassumesthereisadifferenceintheimpactof
terrorismontourismbetweendevelopedanddevelopingcountries.Adummyvariableis
utilizedinthissubsampletodifferbetweendevelopedanddeveloping countries.
Developedcountriesare givenavalueof0anddevelopingcountriesare givenavalueof
1,allowingslopesandinterceptstodiffer.Thedummyvariable,D,isinteractedwiththe
otherindependentvariablesinthemodelasshownbelow.
ATC=B1 +B2PPP+B3TERRi +B4GDPSi +B5GDPHi +B6D+B7(D*PPP)
+B8(D*TERR)+ B9D*GDPSi +B10(G*GDPHi).
ThenullhypothesisoftheChowTestisH0:B6 =B7 =B8 =B9 =B10 =0.Ifthenull
hypothesiscannotberejectedthentheunconstrainedsubsamplereducestothe
constrainedsubsample andthereisnotastatisticallysignificantdifference between
developedanddevelopingcountries?tourismsectors.Ifthenullhypothesiscanbe
rejected,thisprovides furtherevidencethatterrorismdoesaffectdevelopingcountries?
tourismmorethandevelopedcountries?tourismsector.
Totestthenullhypothesisthefollowing Ftestisemployed:

40
F=((SSEC-SSEU)/kC)/(SSEU)/n-ku)
whereSSEUistheresidualsumof squaresfromtheunconstrainedsample,SSECis the
residualsumofsquaresfromtheconstrainedsubsample,kC isthenumberofconstraints
(5),andkuisthenumber ofindependentvariables intheunconstrainedsample,andnis
thenumberofobservations.AstatisticallysignificantFstatisticindicates wepreferthe
unconstrainedmodeltotheconstrainedmodelandrejectthenullhypothesis.Thus,there
isadifferencebetweenthetwosubsamples.
ToutilizetheChowtest onthesampleofcountriesinthethesiscountriesare
dividedintodevelopedanddeveloping countries.Thisinformationcomesfromthe
WorldBank.TheWorldBankclassifiescountriesbasedontheir2006GNI(gross
nationalincome).The WorldBanktakesintoaccountchangesinthe exchangeratesand
inflationratesduring2006aswellastheprevioustwoyearsduring calculation.The
WorldBankclassifies countriesintofourcategories,low,lowermiddle,uppermiddle,
andhighincome.Incomesare expressedin2006U.S.dollars.Lowincomecountriesare
thosewithaGNIlessequaltoorlessthan$905.Lowermiddleincome countriesare
thoselyingbetween$906and$3595.Uppermiddleincomecountrieslie between$3596
and$11,115.HighincomecountrieshaveaGNIequaltoor greaterthan$11,116.For
purposesofthisthesis,countriesinthelowandlowermiddleincomecategoriesare
considereddevelopingcountries.Countriesintheuppermiddleandhighincome
categories are considereddevelopedcountries.
TheresultsfromOLSanalysisindicate alargedifferencebetweendeveloped
anddevelopingcountrieswithregardstoterrorism?seffectonthetouristindustry.We

41
shouldexpectastatisticallysignificantF statistic supporting thestrongereffecton
developing countries.
TheresultsfromtheChowTestrejectthenullhypothesis.TheF-statisticcritical
valueat5%is2.40.The degreesoffreedomforthenumerator anddenominator are5
and50respectively.Thenumeratordegreesoffreedomequalthenumberofrestrictions
andthedenominatordegreesof freedomareobtainedbysubtractingthenumberof
independentvariablesfromthenumberofobservations.Thevalueobtainedfromthe
derivationis2.8.ThenullisrejectedandthereforetheChowtestdoessupportthe
differenceinthe effectofterrorismondevelopedversusdevelopingcountries.
D.TestingforHeteroskedasticity
TheBreuschPaganTest detectsheteroskedasticity.Heteroskedasticityis aresult
ofnonconstantvariance.HeteroskedasticityleadstoinefficiencyofOLSestimateswhich
cannotbetrusted.Thedatasamplecontains countriesthathaveasignificantnumberof
attacks(Columbiahasover1000)andcountrieswithlittleornoattacks.Theseoutliers
couldleadtoheteroskedasticresults.
TheBreuschPaganTest involvesdividingthesumofthesquaredresidualsbythe
numberofobservationsinthesample(SSR/n)denotedbythe Greeklettersigma,?.Pi
(?),thevariable constructedbydividingthesquaredresidualsbysigma,isregressedon
theindependentvariables.Thistestshows whethertheindependentvariablesareinany
waycorrelatedwiththeerrorterm.Then*R2obtainedfromthisregressionfollowsachi-

42
squareddistributionwiththedegreesoffreedomequaltothenumberofindependent
variables.
Thenullhypothesisassumeshomoskedasticityor constantvariance.Thetest
statisticfortheBreuschPagantestisn*R2where(n)isthenumberofvariablesinthe
model(60)multipliedbytheR2valueobtainedbytheBreuschPaganmodel(.122).
Heteroskedasticityisindicatedby a (n*R2)valuelargerthanthecritical ?2value.The
degreesoffreedomisequaltothenumberofvariablesintheregression(5).The ?2
criticalvaluefor5degreesoffreedomis11.07.Theobtainedfromthe appropriate
regressionis7.35,andthedataishomoskedastic.Thehomoskedasticresultsshowthat
varianceisconstantandtheoriginalmodelpasses theBreuschPaganTest.
Anotherstepistotestformisspecificationofthe modelusingtheRamsey Reset
Test.
E.Misspecification
TheRamseyResetTest testsformisspecificationintheregression.Nonlinearity
arisesfrommisspecificationofthemodel.FailuretopasstheRamseyTest requires
reevaluationofthevariablesincludedinthemodel.
ThefirststepoftheResetTestistoruntheregressionE(y/x).Thepredictedy
fromthisregressionisthensquared,cubed,raisedtothefourth,andincludedinthe
regressiony= B1xi+B2YHAT2+B3YHAT3+B4YHAT4. Thenullhypothesisisthen
testedthatH0:B2=B3=B4=0.ThemodelpassestheRamseyTestifthet-statisticforthe
coefficientB2,B3,andB4arenotsignificant.Ifthecoefficientsaresignificantthen

43
presumablythereissomeeffectinthedependentvariableoccurringthattheexplanatory
variablesarenotcapturing.AjointFtestcanbeusedtotestthenullhypothesis.TheF-
statisticcanbederivedfromthefollowingequation:
((R2(new)-R2(old))/r)/((1-R2(new))/(n-k)),
whereR2(new) istheexplainedsumofsquaresfromthemodelincludingthe predictedy
variables,R2(old)istheexplainedsumofsquares fromthemodelexcludingthepredictedy
variables,(r)isthenumberofrestrictions(3),(n)isthenumberofobservations(60)and
(k)isthenumberofindependentvariablesincludingthepredictedyvariables(8).The
derivedF-statisticis0.85.Thenumerator?sdegreesoffreedomareequaltothenumber
ofrestrictions(r)andthe denominator?sdegreesoffreedom are equalto(n-k).The
criticalF-statisticfor3degreesoffreedominthe numeratorand52degreesoffreedomin
thedenominatoris2.78.Thederivedstatisticissmallerthanthecriticalstatistic
(0.85<2.85).Theinsignificanceofthepredictedyvariablesmeansthenullhypothesis
cannotberejectedandthemodeliscorrectlyspecified.
F.Model Comparison
Theinfluenceoftheinteractivetermonthe resultssupportsitsinclusioninthe
originalmodel.Byexcludingtheinteractiveterm andrunningthefollowingregression,
TPC=B1 +B2PPP+B3TERR+B4GDPS +B5GDPH +ei
resultsbetweenthetwomodelsarecompared.Wecanexpectsimilarcoefficient
estimatesandhight-statisticsforTERRandGDPH.TheresultsinTableAshowthese
expectationsarevalid.

44
Terrorism?s effectontourismarrivalsper capitadecreasesfrom -0.22to-0.17inthe
noninteractionmodel.Theeffectofterrorismontouristarrivalsper capita is-0.07for
highGDPHcountriesand-0.37forlowGDPH countries.Asexpected,the coefficient
estimateinthesecondmodel,-0.17,liesinbetween-.07and-.37sincewearenot
interactingthetwovariablesinthesecondmodel.Further,-0.17isquitesimilartothe
coefficientestimateof-0.22formtheoriginalmodel.Recallthat-0.22isthecoefficient
estimatefromtheoriginalmodelbeforethedifferentialeffectofterrorismontourism
betweenhighandlowincomecountries arederived.Thesetwoestimates(-0.17and-
0.22)aresimilar.Thesignificanceoftheinteractivetermintheoriginalmodelsuggests
that-0.22mightbeamorerealisticcoefficientestimatethan-0.17.
ThecoefficientestimateinthenoninteractivemodelforGDPHis1.14comparedto
1.17intheoriginalmodel.Again,thesecoefficientestimatesareverysimilar.The
slightlyhighercoefficientestimateintheoriginal modeldoessuggestthatterrorism
positivelyinfluencesGDPH?s effectontourismarrivalspercapita.Intuitivelythisdoes
notmakemuchsense.I believewhatisimportanttonoteherethanineithercase GDPH
hasaverylarge effectonthetourismindustry.
G.TourismExpenditureModel
Besidestourismarrivals,tourismreceiptsareoftenusedasthedependentvariable
intourismmodels. Receiptsmeasurethe amountthattouristsspendinthetourist
country.Arrivalsandreceiptsarehighlycorrelated,yetdifferencesmayoffer further
insightintotouristtastesandpreferences.

45
Thefollowingmodelusestourismreceiptsasthe dependentvariable.The
independentvariablesareidenticaltotheoriginalmodel.ResultsarelistedinTable8.
Theresultsindicatetheonlyvariableofstatistical significanceisGDPHwithat-statistic
of3.57.ThecoefficientestimateofGDPHis0.97.A1%increaseinGDPHincreases
tourismrevenuesby0.97%.Asexplainedpreviously,highGDPHisanindicationofa
highleveloftourismaccommodationandamenities.Thus,thepositivecorrelation
betweentourismrevenuesandGDPHisexpected.Thiseffectislarge yetGDPH?seffect
ontourismarrivalsper capitaislarger(1.17).Thisdifferenceisunderstandablebecause
whilethehighlevelofcomfortandamenitiesindicatedbyhighGDPHwill attractmore
tourists,theextenttowhichtheyindulgethemselveswhileonvacationdependsontheir
budgets.
Itmustalsobenotedthatthesteepnessofthedemandcurveindicatesthe tourism
industry?spriceelasticity.Elasticitymeasureshowtouristsreacttochangesinpricesata
givenpriceandquantity (?Q/?P*P/Q).Asteeperdemandcurveindicatesaninelastic
demandcurve,anincreaseinpricesdoesnotdeter tourism.Iftourismdemandisprice
inelastic,anincreaseinpricesincreasestourismrevenue.Ata givenprice andquantity,a
flatterdemandcurvemeanstourismdemandiselastic.Iftourismdemandispriceelastic
thenanincreaseinprice willdecreasethequantity oftourismsubstantiallyanddecrease
tourismrevenue.
Thepriceoftourism,PPP,isstatisticallyinsignificantinthismodel.Therefore,
priceoftourismisunitelastic.Unitelasticityimpliesthatanincreaseinthepriceof
tourismbyacertainamountwilldecreasetourismarrivalspercapitabythesameamount.

46
Thetwoaffectsoffseteachotherandtourismrevenuesremainthesameas beforethe
priceincrease.
Itisimportanttonotethedifferencebetweenthequantityoftourismdemandedand
thenumberoftourists.Inthefirstmodel,realpricesdonotaffectTPC(numberof
tourists).Inthethirdmodel,duetounitpriceelasticityofdemandfortourism,an
increaseinpricesisoffsetbyadecreaseinthequantitydemandedfortourism.Although
thequantitydemandedfortourismmaydecrease,thenumberoftouristarrivalsisthe
same.For example,anincreaseintourismprices mayshortentheirlengthofstay,butthe
samenumberoftourists willstillvisitthecountry.
a.BreuschPagan-Model2
TheBreuschPaganteststherevenuemodelforheteroskedasticity.Again,the
conclusionfocusesontheresultsfromthefirstmodelbecausetheR2valueforthatmodel
ismuchhigher.Byregressingthepivariableconstructedfromthismodelonthe
independentvariablesthecriticalnR2valueof8.88isobtained.The ?2 (Chisquared)
criticalvaluefor5degreesoffreedomis11.07.Thevalue8.88islessthanthecritical
valuethereforethemodelishomoskedasticorhasconstantvariance.
b.MisspecificationTestModel2
Asintheprimaryregression,theRamseyResetTestisusedtoindicatemodel
misspecificationinmodel2.Thepredictedy fromregressingtheindependentvariables

47
ontourismreceiptsissquared,cubed,raisedtothefourththenincludedinthefollowing
regression:
Y=B1Xi +B2YHAT2 + B3YHAT3 +B4YHAT4 +ei.
Statisticallysignificantpredictedyt-statisticsindicatemisspecification.Thenull
hypothesisisH0:B2=B3=B4=0.Aspreviously,theF-statisticisderivedtotestthenull
hypothesis.The formulafortheF-statisticis:
((R2(new)-R2(old))/r)/((1-R2(new))/(n-k))
TheF-statisticderivedfromthisequationis1.45.Thecritical Fstatisticat5%
with3degreesoffreedominthenumeratorand52degreesoffreedominthe
denominatoris2.78.Thenullhypothesiscannotberejectedthereforethe modelis
correctlyspecified.

48
TABLE8.ResultsofModels
Model1Model2Model3
DEPENDENT
VARIABLES/MODEL
TPC/INTERACTION TPC/NO INTERACTION REC/INTERACTION
INDEPENDENT
VARIABLES
INTERCEPT -8.47
(-2.33)
-12.46
(-4.05)
10.4
(3.02)
PPP -0.06
(-0.95)
-0.07
(-1.16)
-0.04
(-0.74)
TERR **-1.67
(-2.18)
**-0.17
(-2.25)
-0.75
(-1.04)
GDPS 0.06
0.25
0.06
(0.23)
0.17
(0.73)
GDPH ***0.70
(2.45)
***1.14
(6.37)
***0.97
(3.57)
INTERACTIVE **0.16
(1.96)
0.11
(1.36)
R2 .57 .54 0.61
F 18.23 21.91 21.51
n 60 60 60
T-statisticsareinparenthesis
*significant atthe10%criticalvalue(1.30for55degreesoffreedom)
**significant atthe5%criticalvalue(1.67for55degreesoffreedom)
***significant atthe1%criticalvalue(2.40for55degreesoffreedom)

49
5.CONCLUSION
Therepresentativetouristsappeartopreferqualityvacationsoverbargain
vacations.HighGDPH mayserve asanindicationofqualitytotourists.SpainandItaly
continuetobeamongthetopfivetouristdestinationsdespitetheadoptionoftheeuro
whichiscurrentlyat analltimehighvalue. PPP?sinsignificance andGDPH?s
significancesignifytheprice unitelasticityoftourismdemand.Individual casestudies
mayrevealforsome countriesthattourismispriceinelastic,furtheradvocatingits
developmentandroleinaneconomy.
Manydevelopingcountriesrelyheavilyonagriculturalexports.Agriculturecan
bevulnerabletodroughtsandfluctuationsinworldprices.Tourismdemandisunit
elastic.Investmentinthetourismsectormaycontributetomoreeconomicgrowththan
agriculturalinvestmentwhosedemandismoresensitivetoprices.Economically,it
makesmoresenseforpoorcountriesthatcouldpotentiallyattracttouriststoinvestin
tourismratherthanagriculture.Many charitiesprovideeconomicaidtodeveloping
countriesbysupportingtheiragriculturalindustries.These groupsmayfindtheirmoney
puttobetterusebysupportingthetourismindustry inpoorercountries.Spillovereffects,
likeimprovementstoinfrastructure,may resultfromtourismsectorinvestment.

50
Thisstudyisuniquebyusing crosssectionalanalysistocompareterrorismin
developedanddevelopingcountries.Theconclusionisthatterrorismisparticularly
detrimentaltotourisminundevelopedcountries.Thedevelopedcountries?tourism
industriesareonlyslightly affectedbyterrorism.Economicdiversity andavailabilityof
resourceslessenthe effectofterrorismindevelopedcountries.
Thoughtourismisasectordominatedbydevelopedcountries,itshouldnotbe
overlookedbydevelopingcountries.Tourismcomprises10%ofworldGDPand
continuestogrow.Touristpreferenceforqualityisreflectedbythesignificanceof
GDPH.Developingcountriessuffering fromterrorismfacemanyobstaclesiftheywish
tocultivatetheirtourismindustry.Additionalsecurityisthemosteffectivewayto
combatterrorists.Funds shouldbeappropriatedforsecuritytoestablisha safeimage for
thecountry.Governmentsofdeveloping countriesmustfindwaystoaddressthe security
dilemmatocombatterrorismwithoutreducingfundingtoothereconomicsectors.
Thisthesisprovidesabackgroundfor futurestudiesonthesametopic.The
effectofterrorismontourismisaninterestingandrelevanttopictodaythat warrants
furtherstudy.
Oneareathatthisthesiscouldimproveuponisintheareaofdatacollection.Due
tolimitedresources,tourismsourcedata(country names)camefrom1996whenthebase
yearofthestudy was2003.
Apaneldatastudyconsistingofcountriesinthesamegeographic regionwould
yieldmore concreteorbelievableresults.The crosssectionalanalysisincludesmany
countriesfromvariousregions.Itisdifficulttodeterminewhytouristsare attractedto

51
someareasandnotattractedtootherareas.Furthermore,thecountriesthat dobenefit
fromtourismmay attract touristsfordifferentreasons.Somecountriesmay havenice
beaches,mountainswithgoodhiking,orhistoricsights.Crosssectionalanalysis doesnot
explainwhytouristscontinuetogotothesame countriesornottoothers.
Apaneldatastudyinvolvingcountriesinthesamegeographicareamaybetter
explaintouristpatterns.Presumably,countriesinthesameregionshouldshare
geographictraitsandperhapsevensharesimilartraditionsandcultures.Thus,theeffect
ofaterroristattackontourismwouldbeimmediatelyfelt.Atouristwouldsimply
substituteavacationinonecountrytoitsneighboringcountryiftheysharedsimilartraits
thatattractedthetouristtotheregioninthefirstplace.Theeffectofterrorismontourism
wouldbebetterisolatedifthecountriesinthestudy sharesimilartouristattractions.

52
6.SUMMARY
Terrorismcancausetremendouseconomicshocks.Thetourismsectorisoneof
theeconomicsectorsmostdirectly affectedbyterrorism.Terrorismincidencesdeter
touristarrivals,foreigninvestment,anddestroysinfrastructure.Despitethe damage
terrorismcanhaveontourism,theextenttowhichcountriesareaffectedby terrorism
attacksdependlargelyonthestrengthofitseconomy.Thispaperclearlyshowsthat
developednations?tourismismuchlessaffectedby terrorismthandevelopingcountries?
tourism.Theabilitytofinance addedsecurityisoneofseveralreasons givenbythis
thesisastothedifferenceintourismshocksduetoterrorismbetweenrichandpoor
countries.

53
REFERENCES
1. Aly,HassanY.andMarkC.Strazicick(2000).?TerrorismandTourism: Is the
ImpactPermanentofTransitory?TimeSeriesEvidencefromSomeMENA
Countries?OhioStateUniversity,Departmentof Economics,Columbus,Ohio.
2. Brakke,Michael(2004) ?InternationalTourism,Demand,andGDP Implications:A
BackgroundandImpiricalAnalysis,? UndergraduateEconomicReview (1).
3. CompendiumofTourismStatistics2003
4. Drakos,KonstantinosandAliM.Kutan(2003)?RegionalEffectsofTerrorismon
Tourism:Evidencefrom3MediterraneanCountries,? JournalofConflictResolution
45(5).;pp.621-64.
5. Enders,WaltandToddSandler(2003)?Aneconomicperspectiveontransnational
terrorism.?
6. Enders,WaltandToddSandler,?EconomicConsequencesofTerrorismin
DevelopedandDevelopingCountries:AnOverview,?2005.
7. Enders,WaltToddSandlerandGeraldParise(1992) ?AnEconometricAnalysisof
the ImpactofTerrorismonTourism,? KYKLOS45.;pp.531-554
8. Garin-Munoz,T.andPerezAmaral,T.(2000) ?Aneconometricmodel for
internationalflowstoSpain.?AppliedEconomics(7).;pp.525-529.

54
9. Gujarati,DamodarN.(2003),BasicEconometrics,McGraw-Hill,Singapore.
10. Henderson,JoanC.(2003),?TerrorismandTourism:ManagingtheConsequencesof
theBali Bombings,? JournalofTravelandTourismMarketing 15(1).
11. Looney,Robert(2002) ?EconomicCoststotheUnitedStatesStemmingformthe
9/11Attacks,? StrategicInsights 1(6).
12. PennWorldTable,http://pwt.econ.upenn.edu/
13. Sonmez,Sevil,YiorgosApostolopoulosandPeterTarlow(1999)?TourisminCrisis:
ManagingtheEffectsof Terrorism,? JournalofTravelResearch.
14. TourismMarketTrends,Europe2004,WTO
15. YearbookofTourismStatistics2004
16. Schmidt,Stephen(2005),Econometrics,McGraw-Hill/Irwin,NewYork,NY.

55
REGRESSIONANALYSISRESULTSUSINGLIMDEP
MODEL1: Lhs=LNARRPCA;Rhs=ONE,LNPPP,LNTERR,LNGDPSOU,LNGDPHOS,INTER$
| Ordinary least squares regression |
| Model was estimated Apr 06, 2008 at 00:55:11PM |
| LHS=LNARRPCA Mean = -2.309327 |
| Standard deviation = 1.779687 |
| WTS=none Number of observs. = 60 |
| Model size Parameters = 6 |
| Degrees of freedom = 54 |
| Residuals Sum of squares = 79.86644 |
| Standard error of e = 1.216145 |
| Fit R-squared = .5726093 |
| Adjusted R-squared = .5330361 |
| Model test F[ 5, 54] (prob) = 14.47 (.0000) |
| Diagnostic Log likelihood = -93.71665 |
| Restricted(b=0) = -119.2184 |
| Chi-sq [ 5] (prob) = 51.00 (.0000) |
| Info criter. LogAmemiya Prd. Crt. = .4866819 |
| Akaike Info. Criter. = .4860112 |
| Autocorrel Durbin-Watson Stat. = 1.8706852 |
| Rho = cor[e,e(-1)] = .0646574 |
+----------------------------------------------------+
+---------+--------------+----------------+--------+---------+----------
+|Variable | Coefficient | Standard Error |t-ratio |P[|T|>t] | Mean X
|+---------+--------------+----------------+--------+---------+---------
-+
Constant -8.46528522 3.63176445 -2.331 .0235
LNPPP -.05593343 .05875897 -.952 .3454 2.55436943
LNTERR -1.66656932 .76547635 -2.177 .0339 2.95446854
LNGDPSOU .06058853 .24171834 .251 .8030 9.94085881
LNGDPHOS .70022444 .28589920 2.449 .0176 8.98085317
INTER .16423841 .08380972 1.960 .0552 26.3747344

56
MODEL 2: Lhs=LNARRPCA;Rhs=ONE,LNPPP,LNTERR,LNGDPSOU,LNGDPHOS$
+----------------------------------------------------+
| Ordinary least squares regression |
| Model was estimated Apr 06, 2008 at 01:00:32PM |
| LHS=LNARRPCA Mean = -2.309327 |
| Standard deviation = 1.779687 |
| WTS=none Number of observs. = 60 |
| Model size Parameters = 5 |
| Degrees of freedom = 55 |
| Residuals Sum of squares = 85.54622 |
| Standard error of e = 1.247151 |
| Fit R-squared = .5422151 |
| Adjusted R-squared = .5089216 |
| Model test F[ 4, 55] (prob) = 16.29 (.0000) |
| Diagnostic Log likelihood = -95.77768 |
| Restricted(b=0) = -119.2184 |
| Chi-sq [ 4] (prob) = 46.88 (.0000) |
| Info criter. LogAmemiya Prd. Crt. = .5217663 |
| Akaike Info. Criter. = .5213789 |
| Autocorrel Durbin-Watson Stat. = 1.8759510 |
| Rho = cor[e,e(-1)] = .0620245 |
+----------------------------------------------------+
+---------+--------------+----------------+--------+---------+----------
+
|Variable | Coefficient | Standard Error |t-ratio |P[|T|>t] | Mean X|
+---------+--------------+----------------+--------+---------+----------
+
Constant -12.4637141 3.08102989 -4.045 .0002
LNPPP -.06945223 .05984034 -1.161 .2508 2.55436943
LNTERR -.17376427 .07717495 -2.252 .0284 2.95446854
LNGDPSOU .05823170 .24787802 .235 .8151 9.94085881
LNGDPHOS 1.14313210 .17955432 6.366 .0000 8.98085317

57
MODEL 3: Lhs=LNREC;Rhs=ONE,LNPPP,LNTERR,LNGDPSOU,LNGDPHOS,INTER$
+----------------------------------------------------+
| Ordinary least squares regression |
| Model was estimated Apr 06, 2008 at 01:08:00PM |
| LHS=LNREC Mean = 21.27961 |
| Standard deviation = 1.760484 |
| WTS=none Number of observs. = 60 |
| Model size Parameters = 6 |
| Degrees of freedom = 54 |
| Residuals Sum of squares = 71.82402 |
| Standard error of e = 1.153289 |
| Fit R-squared = .6072163 |
| Adjusted R-squared = .5708474 |
| Model test F[ 5, 54] (prob) = 16.70 (.0000) |
| Diagnostic Log likelihood = -90.53254 |
| Restricted(b=0) = -118.5674 |
| Chi-sq [ 5] (prob) = 56.07 (.0000) |
| Info criter. LogAmemiya Prd. Crt. = .3805450 |
| Akaike Info. Criter. = .3798744 |
| Autocorrel Durbin-Watson Stat. = 2.0466080 |
| Rho = cor[e,e(-1)] = -.0233040 |
+----------------------------------------------------+
+---------+--------------+----------------+--------+---------+----------
+
|Variable | Coefficient | Standard Error |t-ratio |P[|T|>t] | Mean X|
+---------+--------------+----------------+--------+---------+--------+
Constant 10.3967532 3.44405716 3.019 .0039
LNPPP -.04129735 .05572202 -.741 .4618 2.55436943
LNTERR -.75146558 .72591280 -1.035 .3052 2.95446854
LNGDPSOU .16815387 .22922516 .734 .4664 9.94085881
LNGDPHOS .96825951 .27112253 3.571 .0008 8.98085317
INTER .10772173 .07947802 1.355 .1809 26.3747344

58
DATAAPPENDIX
VARIABLESUMMARY
*Allvariablesareinnaturallog form
LNARRPCAP-touristarrivalspercapita
lnrec-totaltourismreceipts
lnppp-purchasingpowerparity (priceoftourism)
lnterr-numberofterroristincidences
lngdpsource-GDPpercapitaofthetouristsource country
lngdphost-GDPpercapitaofthetouristhostcountry
inter-variableinteracting lnterrwithlngdphost
dev.-developingcountry dummy
int-variableinteractinglnterrwithdev.
COUNTRIES LNARRPCAP lnrec lnppp lnterr lngdpsource lngdphost Inter Dev
ARGENTINA -2.55995 21.46377 -0.16252 2.944439 9.152212 9.344738 27.52 0
BRAZIL -3.7954 21.63112 0.09531 2.484907 9.344738 8.962003 22.27 0
CHILE -2.27273 20.57244 5.497865 2.397895 9.344738 9.492733 22.76 0
CHINA -3.66445 23.58008 0.683097 2.397895 10.14747 8.579467 20.57 1
COLOMBIA -4.2012 20.51992 6.740968 7.034388 8.839364 8.773125 61.71 1
COSTA
RICA -1.14567 20.87255 5.303802 0 10.52711 9.093445 0 0
CYPRUS 1.093424 21.42389 -0.99425 2.302585 10.24359 10.07295 23.19 0

x
EGYPT -2.56523 22.24584 0.223144 2.397895 10.20693 8.536811 20.47 1
INDONESIA -3.96241 22.11877 7.49527 5.283204 10.2329 8.349229 44.11 1
JORDAN -1.24451 20.5187 -1.20397 2.484907 8.536811 8.372877 20.81 1
LEBANON -1.3009 20.73914 6.957887 4.234107 8.372877 8.86695 37.54 0
MALAYSIA -0.78085 22.49839 0.24686 1.386294 10.2329 9.496844 13.17 0
MEXICO -1.72151 22.97002 2.052841 2.70805 10.52711 9.04595 24.5 0
PHILLIPINES -3.79264 21.10444 2.504709 5.361292 10.52711 8.274339 44.36 1
THAILAND -1.85227 22.78021 2.479894 6.428105 9.496844 8.944868 57.5 1
Australia -1.51095 23.05716 0.329304 0 10.14747 10.32846 0 0
BELGIUM -0.43916 22.81883 -0.0202 2.484907 10.24359 10.20514 25.36 0
CZECH. REP -0.70155 21.80981 2.721295 1.098612 10.20693 9.706502 10.66 0
GERMANY -1.50189 23.85806 -0.03046 2.70805 10.24908 10.20693 27.64 0
GREECE 0.232165 23.0936 -0.19845 5.438079 10.20693 9.738999 52.96 0
HUNGARY 0.436543 21.95874 4.859657 0 10.20693 9.564264 0 0
IRELAND 0.463482 22.07781 0.173953 1.609438 10.24359 10.28869 16.56 0
ISRAEL -1.7499 21.43573 1.358409 6.137727 10.52711 9.997936 61.36 0
ITALY -0.38083 24.16439 -0.09431 4.828314 10.20693 10.11381 48.83 0
JAPAN -3.199 22.90346 5.027624 2.995732 9.806871 10.14747 30.4 0
NORWAY -0.33621 21.65582 2.300583 0 10.20693 10.44953 0 0
PORTUGAL 0.119688 22.66014 -0.41552 0 10.02262 9.848796 0 0
RUSSIA -2.89215 22.22779 2.013569 5.648974 8.837577 9.410634 53.16 0
SLOVENIA -0.34346 21.01743 4.917862 0 10.20693 9.985923 0 0
S. AFRICA -1.86269 22.17488 1.118415 3.135494 7.781001 9.172628 28.76 0
SPAIN 0.206874 24.45544 -0.19845 6.595781 10.24359 10.02262 66.11 0
U.K. -0.8749 23.84792 -0.41552 3.218876 10.2275 10.24359 32.97 0
U.S. -1.96013 24.89007 0 4.584967 10.30145 10.52711 48.27 0
INDIA -5.95342 21.98229 2.098018 6.232448 10.24359 8.074814 50.33 1
PAKISTAN -5.75131 18.603 2.479894 6.222576 10.24359 7.893538 49.12 1
SRI LANKA -3.67392 19.86524 2.876949 4.875197 10.24359 8.506735 41.47 1
BANGLAD. -6.33699 17.85856 2.327278 4.70953 10.2338 7.762698 36.56 1
TURKEY -1.67539 23.30371 13.65651 6.167516 10.20693 8.683008 53.55 0
NEPAL -4.36071 19.10882 2.395164 5.937536 10.2338 7.337698 43.57 1
ALGERIA -3.33741 18.89691 3.24142 4.143135 10.2275 8.701431 36.05 1
PERU -3.41823 20.64314 0.470004 3.258097 10.52711 8.442849 27.51 1
IRAN -3.81813 21.29819 7.687993 3.688879 7.893538 8.882685 32.77 1

xi
MOROCCO -1.94041 21.77269 1.105257 1.791759 10.2275 8.384121 15.02 1
VENEZUELA -4.29264 19.59316 6.699414 4.382027 10.52711 8.839364 38.73 0
HONDURAS -2.39025 19.63559 1.993339 1.609438 10.52711 7.785771 12.53 1
FRANCE 0.223648 24.33521 -0.04082 6.100319 10.20693 10.2275 62.39 0
TUNISIA -0.66305 21.18259 -0.91629 0 10.20693 8.997366 0 1
KENYA -3.5302 19.64151 3.282789 1.609438 10.20693 7.165911 11.53 1
ECUADOR -2.89126 19.82186 -0.84397 3.637586 8.773125 8.43427 30.68 1
CAMBODIA -3.82743 19.77909 6.732973 2.564949 9.937918 6.423604 16.48 1
GUATEMA -2.7604 20.24684 1.22083 1.098612 10.52711 8.315559 9.14 1
ALBANIA -4.45453 20.07318 3.804883 2.397895 10.11381 8.447985 20.26 1
LAOS -3.40824 18.28142 7.758816 2.079442 8.944868 7.32138 15.22 1
NICARAGUA -2.27727 18.83279 1.226712 0 10.52711 8.180228 0 1
COTE D IV'. -4.5458 18.24633 5.241377 0 10.2275 7.857558 0 1
N. ZEALAND -0.62885 22.10304 0.371564 0.693147 10.32846 10.09746 6.99 0
EL
SALVADOR -2.02155 19.31677 1.386294 0.693147 10.52711 8.527013 5.910 1
BELARUS -5.08316 19.40276 5.573674 0 9.410634 9.518281 0 1
URUGUAY -0.87703 19.65905 2.282382 1.098612 9.344738 9.152212 10.05 0
MADAGAS. -4.8053 18.14624 7.781042 1.94591 10.2275 6.724806 13.09 1