Upload
sportskenya-tm
View
219
Download
0
Embed Size (px)
Citation preview
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
1/30
TheOlympicEffectAndrewK.RoseandMarkM.Spiegel*
May14,2010
AbstractEconomistsareskepticalabouttheeconomicbenefitsofhostingmegaeventssuchastheOlympic
GamesortheWorldCup,sincesuchactivitieshaveconsiderablecostandseemtoyieldfewtangible
benefits. Thesedoubtsarerarelysharedbypolicymakersandthepopulation,whoaretypicallyquite
enthusiasticaboutsuchspectacles. Inthispaper,wereconcilethesepositionsbyexaminingthe
economicimpactofhostingmegaeventsliketheOlympics;wefocusontrade.Usingavarietyoftrade
models,weshowthathostingamegaeventliketheOlympicshasapositiveimpactonnationalexports.
Thiseffectisstatisticallyrobust,permanent,andlarge;tradeisover20%higherforcountriesthathave
hostedtheOlympics. Interestinglyhowever,wealsofindthatunsuccessfulbidstohosttheOlympics
haveasimilarpositiveimpactonexports. WeconcludethattheOlympiceffectontradeisattributable
tothesignalacountrysendswhenbiddingtohostthegames,ratherthantheactofactuallyholdinga
megaevent. Wedevelopapoliticaleconomymodelthatformalizesthisidea,andderivesthe
conditionsunderwhichasignallikethisisusedbycountrieswishingtoliberalize.
JELClassification:F19,L83
Keywords:trade,liberalization,megaevent,host,candidate,export;model,signaling,data
AndrewK. Rose(correspondence) MarkM. Spiegel
HaasSchoolofBusiness FederalReserveBankofSanFrancisco
UniversityofCalifornia 101MarketSt.
Berkeley,CAUSA947201900 SanFranciscoCA94105
Tel:(510)6426609 Tel:(415)9743241Fax:(510)6424700 Fax:(415)9742168
Email:[email protected] Email:[email protected]
* RoseisB.T. RoccaJr. ProfessorofInternationalTradeandEconomicAnalysisandPolicyintheHaas
SchoolofBusinessattheUniversityofCalifornia,Berkeley,NBERresearchassociateandCEPRresearch
fellow. SpiegelisVicePresident,EconomicResearch,FederalReserveBankofSanFrancisco. Wethank
ChristopherCandelariaforexcellentresearchassistance,andKeithHeadforhelpwithhistetradic
programs. Forcomments,wethank:AntonioCiccone,JohnFernald;ReuvenGlick;YuriGorodnichenko;PierreOlivierGourinchas;GalinaHale;DennisNovy;AssafRazin;GlennRudebusch;JayShambaugh;
BentSorensen;EricSwanson;LindaTesar;ShangJinWei;JohnWilliams;EricvanWincoop;two
anonymousreferees;andworkshopparticipantsatBerkeley,Copenhagen,FRBSF,NBER,Tsinghua,and
UCSD. Moreextensiveearlierandcurrentversionsofthispaperarepostedat
http://faculty.haas.berkeley.edu/arose,alongwiththerelevantdatasetsandsampleoutput. Theviews
presentedinthispaperarethoseofauthorsanddonotrepresenttheviewsoftheFederalReserveBank
ofSanFranciscoortheBoardofGovernorsoftheFederalReserve.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
2/30
1
WhentheOlympicflameislit,Chinawillbehopingfora17dayfestivalofsportandinternational
friendship. Itseesthegamesasmarkingnotjustitsreemergenceasaglobaleconomicforcebutalsoas
acountrythattherestoftheworldtreatswithadmirationandrespect.
Economist,August2,2008
1. Motivation
Economistsareusuallyskepticalofargumentsaboutthepublicprovisionofinfrastructurefor
sportingevents,andrightlyso. Agentsthatendorsetheconstructionofnewsportsstadiaorthestaging
ofmegaeventsusuallydosooutofnaivetyorselfinterest. Inpractice,theseeventsusuallyendup
imposinglargecostsontheirhoststhatarenotnearlycompensatedbyeithertherevenuesearned
duringtheeventorthelegacyoflargestadiaorobscurefacilities(velodromes,aquaticcenters,andso
forth)thatareleftbehind. Theopeningceremoniesofthe2008BeijingOlympicgamesarecommonly
thoughttohavecostatleast$100millionwhenover100millionChineselivedonlessthan$1/day.
Despitethefactthatmosteconomistsdoubtthewisdomofsuchpolicy,thereislittlequestion
thatcountriescommitsubstantialresourcestobecomecandidatestohostmegaevents,andmuch
moreshouldtheybefortunate"enoughtoactuallyhosttheevent.Whilethereisusuallyavocal
minorityofopponents,thedesiretohosttheOlympicgamesiswidelyheldbyboththemassesand
politicalelites.
Thispaperexaminesthepossibilitythatbothsidesoftheargumentmayberight. Inparticular,
weshowthatthereisalargeeconomicbenefitassociatedwithmegaevents(justifyingthepublic's
enthusiasm),despitethefactthatmuchoftherequisitenewinfrastructureisanetcost(explainingthe
skepticismofeconomists). Succinctly,whenacountrywishestoentertheworldstage,itcansignalthis
tobothdomesticandinternational constituenciesbyofferingtohostamegaevent. Wemodelthis
narrowlybyfocusingontradeliberalization,butbelievethatthepointmaybemoregeneral.
ItiscommonlyarguedthathostingtheOlympicswillpromoteanation'sexports. Forinstance,
theInternationalOlympicCommittee(IOC)believesthatpotentialvisitorswillbedrawntoOlympic
venuesafterbeingexposedtothemthroughthegames. Preuss(2004)discusseshowtheSeoulGames
in1988wereintendedtoimproveinternationalrelationsbetweenSouthKoreaandtheSovietBloc
countriesaswellasraiseinternationalawarenessofKoreanmanufacturedproductssoastopromote
Koreanexports. Wearedubiousofthepracticalrelevanceofthisargument,andthusbeginby
examiningitempirically,usinganumberofdifferentmodelsoftrade. Surprisingly,wefindstrongevidenceofalargepersistenteffectoftheOlympicsonbothexportsandoveralltrade. Ourresults
seemrobusttoabatteryofsensitivityexercisesincludingtetradicandmatching/treatmentestimation;
weshowthatcountrieswhichhavehostedthegamesseemtohaveexportsover20%higher,ceterisparibus. OthermegaeventssuchastheWorldCupalsohavelargepositiveeffectsontrade. Somewhat
surprisingly(atleasttous),countriesthathostamegaeventseemtorealizeaneconomicbenefitinthe
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
3/30
2
formofgreateropenness. Ifopennessenhancesgrowth(e.g.,(Lucas,2009),thewelfareconsequences
ofmegaeventsmaybepositiveandlarge.
Whileourobservedeffectislargeandrobust,itmaybearguedthatitisattributableto
unobservabledifferencesbetweenthosecountriesthathostthegamesandthosethatdon't,rather
thanahostingeffect. Toaddressthispossibility,wecomparetradepatternsforcountriesthathostthegamestothosethatbidunsuccessfullyforthegames. Inusingthisalternativegroupasacontrol,
ourmethodologyfollowsalargerecentliterature,suchasGreenstoneetal(2008). Inourcase,the
operationalassumptionisthatsuccessfulandunsuccessfulcandidatesaresimilarintermsofproclivity
towardstradeandliberalization,differingonlybytheexperienceofactuallyhostingtheOlympicgames.
Usingthisidentificationstrategy,wefindthatcountriesthatwereunsuccessfulcandidatesfor
thegamesalsoexperienceapositiveexporteffect,onesimilarinsizetothatexperiencedbyactual
Olympichosts. Thesefindingscastdoubtontheideathataplausiblemotivationforhostingamega
eventisanychangeinacountry'sfundamentalsinducedbyholdingthegames(suchasconstruction
activityortheresultinginfrastructure). Instead,ourevidenceisconsistentwiththeconclusionthatallcountriesthatbidforthegamesexperienceanincreaseinoutwardorientation,notjusthosts.This
raisestwopuzzles. WhyshouldbiddingfortheOlympicsbeassociatedwithincreasedopenness? Andif
hostingthegamesbringsnotangiblebenefitsrelativetoacontrolgroupofunsuccessfulcandidates,
whywouldanycountryeverbidtoholdanexpensivemegaevent?
Weofferananswertothesepuzzlesbelow. Weexplorethepossibilitythatbiddingtohostan
internationalmegaeventsuchastheOlympicsispartofacostlystrategythatsignalstradeliberalization
andresultsinincreasedopenness. Wedevelopatheoreticalpoliticaleconomymodelconsistentwith
thisconjecture.Inthemodel,weobtainaseparatingequilibriumwherebiddingtohostamegaevent
providesapositivesignalaboutfuturepolicyintentions. However,consistentwithourempirical
findings,hostingthegamesinandofitselfhasnoimpactonanation'sfundamentalsortrade.
OurmodelalsosuggestsananswertothequestionWhyamegaevent?"asthechoiceofa
signalofliberalizationintentions.WefindthatcountrieswillbemorelikelytousetheOlympicsignal,
thegreateristheincidenceoftheexpectedcostofsendingthesignalonthegroupthatexpectsto
benefitfromfutureliberalization. MajorsportingeventsliketheOlympicsaretraditionallyfinancedby
therelevantcity(usuallythecapital)inconjunctionwiththecentralgovernmentofthehostcountry.
Policymakersfromthesegroupsarelikelytobenefitfromliberalization.Assuch,biddingforthe
Olympicsmayservewellasasignalbecauseitalignsthecostsandbenefitsofthesignal. Toour
knowledge,thisisthefirstmodelwherethedistributionalimplicationsofsendingthesignalinfluence
thedesirabilityofthesignalchosen. Olympicbidsarealsogoodsignalsofliberalizationbecausethey
arehighlyvisible,infrequent,andhavelongleadtimes.
2. LiteratureReview
2.1 MegaEvents
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
4/30
3
Anumberofstudiessupportlocalsubsidizationofsportingevents;theyoftenpredictlarge
economicbenefits.Thisworkistypicallycommissionedandisnotintendedfortheacademicaudience,
butitisstillinfluential.ForexampleHumphreysandPlummer(1995)estimatetheshorttermeconomic
impacttoAtlantafromhostingthe1996gamestobe$5.1billion.
Morerigorousstudiesareskepticaloftheneteconomicbenefitsofhostingmegaevents;seee.g.,Owen(2005). Thecostsofholdingsucheventsseemconsiderable. Further,anyenduringbenefits
derivemostlyfrominfrastructureinvestmentsthatthehostcitycouldchoosetomakeindependentlyof
thegames. Muchofthespendingontheeventbylocalcitizensisasubstitutefromadifferentleisure
activityorconsumptiongood,ratherthantrueadditionalspending. Moreover,theprojectsassociated
withthegamestypicallyseemtobewhiteelephants,suchaspoorlyusedsportingfacilitiesassociated
withidiosyncraticOlympicsports,orhotelsandtransportationinfrastructurebuilttoaccommodatea
onetimepeakdemandofjustthreeweeks.
Somehavearguedthathostingsportingeventsyieldsanonpecuniaryfeelgood"benefitto
localcitizenswhoarefilledwithcivicpridefollowingamegaevent,eveniftheydonotattend.
However,theveryexistenceofthisintangiblespilloverisuncertain,letaloneitsmagnitude. Itseems
safetosaythatamajorityoftheprofessionconsidersitunlikelythatthesebenefitsjustifythelarge
publicexpendituresinvolvedinhostingsuchevents(e.g.Coates,2007).
2.2 InternationalSignaling
Theuseofinternationalsignalstoindicatefuturepolicytoprospectiveforeigninvestorshas
beendiscussedintheliterature. BartoliniandDrazen(1997)developamodelwheregovernmentswith
asymmetricinformationaboutfuturefiscalpositionssignaltheirexpectationsthroughcurrentpolicies
oncapitalaccountopenness. Opencapitalaccountsaremorecostlyforcountriesinpoorfiscal
condition,sothosecountriesthatexpecttobeingoodfiscalconditioninthefuturecansignaltheirprospectsinwaysthatcannotbeprofitablymimickedbycountriesthatexpecttofacefuturefiscal
difficulties.
Thesignalweconsiderbelowisofaburningmoney"type,notinformativeinitsownsense,but
informativeduetothefactthatsendingthesignalisonlyattractivetothesetofcountriesthatsincerely
intendtopursueliberalization. Inthissense,itissimilartothecostsofdelayinawarofattritionmodel,
suchasthatofthedelayedfiscalstabilizationinAlesinaandDrazen(1991). Otherstudiesgoeven
farther,e.g.Krugman(1998)andMukand(2006),whoarguethatcountriessometimespursuepolicies
thatareactuallyperverse,inanefforttoincreaseinvestorconfidence. Theirargumentisthatina
globalizedenvironment,policymakersmayfeeltheneedtopursuepoliciesthatwouldconfirmforeigninvestors'beliefsaboutwhatconstitutesgoodpolicy. Thesebeliefsmaybebiasedduetoherding
effects(e.g.Banerjee,1992)oralternativelybecauseinvestorsmaydrawincorrectinferencesabout
theirimpact. Asaresult,perversepoliciesmayhavesuchanadvantageintermsoftheirimpacton
agents'expectations,thatpursuingthemmaybesuperiortofollowingthepaththatwouldyieldthe
bestoutcomeintermsofdomesticeconomicfundamentals.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
5/30
4
Onequestionthatnaturallyarisesinthesetypesofmodelsiswhyoneformofsignalmightbe
preferredtoanother. WearguethatoneattributeofusingtheOlympicsasasignalistheincidenceof
thecostofsendingthesignalwithinthecountry. Inparticular,itislikelythatthecostofhostingthe
Olympicsisprimarilybornebythebenefactorsofthesignaledpolicychange,limitingthelossestothose
notfavoredbythepolicy. MegaeventsliketheOlympicsarealsoinfrequent,highlyvisible,andhave
longleadtimes,attributesthatlendthemselvestosignalsofliberalization.
3. TheOlympicEffectonTrade
Ifthedirecteconomicbenefitsseemtheoreticallydubious,andanyindirecteffectshighly
uncertain,thewillingnessoflocalandfederalgovernmentstoheavilysubsidizesportingactivitiesisa
mystery. Wenowtrytotacklethisissueempirically. Inparticular,wetakeseriouslytheargumentthat
hostingamegaeventprovidesvisibilitytoahostcountryandthusmaystimulateglobaldemandforits
exports.
3.1 SpecificationandData
Westartourinvestigationbyusingthewellknownandwidelyemployedgravity"modelof
internationaltrade(laterbelow,wecorroborateourresultsusingavarietyofdifferentapproaches. This
modelsbilateraltradeflowsbetweenapairofcountriesasafunctionofthedistancebetweenthetwo
countriesandtheireconomicmasses." Weaugmentthisempiricalspecificationbyaddingahostof
otherfactorsthatmightalsoaffecttheirtradeintensity. Weemploythefollowingspecification:
ln(Xijt)= 0+ 1ln(Dij)+ 2ln(Popit)+ 3ln(Popjt)+ 4ln(GDPpcit)+ 5ln(GDPpcjt)
+ 6CUijt+ 7Langij+ 8RTAijt+ 9Borderij+ 10Islandsij+ 11Areaij (1)
+ 12ComColij+ 13Colonyijt+ 14EverColij+ 15SameCtryijt
+ Summerit+ ijt
whereidenotestheexportingcountry,jdenotestheimporter,tdenotestime,ln(.)denotesthenatural
logarithmoperator,andthevariablesaredefinedas:
XijtdenotesrealFOBexportsfromitoj,measuredinmillionsofdollars,
Disthedistancebetweeniandj,
Popispopulation,
GDPpcisannualrealGDPpercapita,
Contisabinaryvariablewhichisunityifiandjsharealandborder,
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
6/30
5
CUisabinarydummyvariablewhichisunityifiandjusethesamecurrencyattimetandzero
otherwise,
Langisabinaryvariablewhichisunityifiandjhaveacommonlanguage,
RTAisabinaryvariablewhichisunityifiandjhavearegionaltradeagreementatt,
Borderisabinaryvariablewhichisunityifiandjsharealandborder,
Islandisthenumberofislandcountriesinthepair(0/1/2),
Areaisthelogoftheproductoftheareasofthecountries,
ComColisabinaryvariablewhichisunityifiandjwerebothcolonizedbythesamecountry,
Colonyisabinaryvariablewhichisunityificolonizesjattimet(orviceversa),
EverColisabinaryvariablewhichisunityifievercolonizedj(orviceversa),
SameCtryisabinaryvariablewhichisunityifiispartofthesamecountryattimet(orviceversa),
isavectorofnuisancecoefficients,
SummerisabinaryvariablewhichisunityifihostedapostwarSummergamesatorbeforetimet,
andzerootherwise,
ijrepresentstheomittedotherinfluencesonbilateralexports,assumedtobewellbehaved.
Thisdatasetincludesannualobservationsbetween1950and2006(thoughwithmanymissing
observations)forsome196territoriesandlocalities(werefertotheseascountries"below). The
sourcesofourbilateraldataset,alongwithacompletelistofthecountriesthemselvesaretabulatedin
thelongerversionofthispaper.
WeestimatethisequationwithOLS,usingarobustcovarianceestimator(clusteredbycountry
pairdyads)tohandleheteroskedasticity. Throughout,weaddyearspecificfixedeffectssoastotake
accountofanytimespecificcommontrendsoreffects. Wealsoperturbthisspecificationintwo
importantways. First,weaddacomprehensivesetofdyadicspecificfixedeffects(i.e.,amutually
exclusiveandjointlyexhaustivesetof{ij}intercepts)toabsorbanytimeinvariantcharacteristicsthat
arecommontoapairofcountries. Second,weaddcomprehensivesetsofexporterandimporterfixed
effects(i.e.,setsof{i}and{j})totakeaccountofanytimeinvariantcountryspecificfactors.1 Wealso
showbelowthatourkeyresultsareinsensitivetotheuseofotherestimationstrategies.
Thecoefficientofinteresttousis . ThisrepresentsthepermanentexporteffecttheOlympic
EffectassociatedwithhavinghostedthepostwarSummerOlympicgames,holdingotherexport
determinantsconstantthroughthegravitymodel(wetryhardnottointerpretthesecoefficientsas
causal). ThehostsofpostwarOlympicgamesaretabulatedinAppendixTableA1.2
3.2 BenchmarkResults
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
7/30
6
TheresultsofestimatingourdefaultspecificationarepresentedinTable1. Thedefault
estimatesaretabulatedinthethreecolumnstowardstheleft,whereresultsarepresentedforthree
differentsetsoffixedeffects(none,dyadic,andexporter/importer).
Beforewediscussthecoefficientofgreatestinteresttous,webrieflydiscusstheother
determinantsoftradeflows.Thegravitymodelseemstoworkwell. Itdeliverspreciselyestimatedcoefficientsthataresensibleandsimilartothoseestimatedbyothers. Forinstance, 1isconsistently
estimatedtobeeconomicallyandstatisticallysignificant;exportsbetweenapairofcountriesfallwith
distance(aspreviousresearchershavefound). Similarly,thesizeandsignificanceof3and 5indicates
thatlargerandrichercountriesbothtendtoimportmore. Exportsarelargerwhencountriessharea
money,language,tradeagreement,landborder,orcolonialheritage. Further,theequationsfitthe
(largelycrosssectional)datasetwell,explainingwelloverhalfofthevariationinexports. Whilethese
resultsarenotofdirectinteresttous,theydoreassureusthatourestimatesaregroundedina
statisticalconditioningmodelthatdeliverssensibleandsignificantresults.
Oncethestandardtradedeterminantsareaccountedforbythegravitymodel,isthereany
roomleftforapermanentexporteffectofhostingtheSummerOlympics? Somewhatsurprisingly(at
leastinitiallytous),theanswerisunambiguouslypositive. Considerthecolumnontheextremeleft
handsideofTable1;theestimateof ispositiveandstatisticallydistinguishablefromzeroatall
reasonablesignificancelevels. Further,theOlympicEffect"iseconomicallylarge;thepointestimateof
.33,takenliterally,impliesthatcountriesthathavehostedthesummergameshaveexportsthatare
permanentlyhigherbysome(exp(.33) 10= 39%! Thisisbig,broadlycomparableinourestimatesto,
e.g.,theeffectsofasharedregionaltradeagreement. Reassuringly,includingeitherdyadicorcountry
specificfixedeffectsdoesnotchangethekeyresultverymuch.
Thefourcolumnsattherightofthetableprovidefurthersensitivityanalysis,adding
combinationsofcountryspecificfixedeffectsinteractedwithlineartimetrends. Thisallows,e.g.,each
exporter(importer)tohaveitsowntrendgrowthrateofexports(imports),asidefromanyotherglobal
timeorothereffects. Thesmallestestimateof appearswhenweincludeexporterspecifictimetrends
andimporterspecificintercepts,buteveninthiscase,previoushostsoftheSummergameshave
exportssome(exp(.16)1=)17%larger.
Succinctly,countriesthathavehostedtheSummerOlympicgamesinthepastseemtohave
exportsthatarebotheconomicallyandstatisticallysignificanthigher,holdingotherthingsconstant;an
effectof20%seemsareasonablelowerbound. Understandingthisapparentlyimplausibleresultmore
deeplytakesuptheremainderofthepaper.
4. SensitivityAnalysis
4.1 RobustnessofPermanentEffect
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
8/30
7
Arethe(surprisingly)stronglinkagesbetweenexportsandOlympichostsfragile? Orinstead,do
smallperturbationstotheexactsampleorchoiceofconditioningmodelhaveastrongeffectonthe
results? No. WeprovideabatteryofrobustnesschecksinTable2,whichisintendedtoreassurethe
skepticalreaderthatourresultsareessentiallyinsensitivetominorchangesintheexacteconometric
methodologyusedtoestimate . Eachoftherowsinthetablecorrespondstoadifferentsensitivity
check,whilethethreecolumnscorrespondtothedefaultthreesetsoffixedeffectsincludingtotheleft
inTable1. Weonlyreportestimatesforthecoefficientofinterest ;othercontrolsareincludedinthe
regressionsasappropriatebutnotreported.
ThefirstexperimentofTable2showstheresultswhenexportsfromitojarereplacedby
importsintoifromj. Thecoefficientsremainstatisticallylargeandpositive. Indeed,thepointestimates
are,ifanything,somewhatlargerthanthoseassociatedwithexports. ThisindicatesthattheOlympics
areassociatedwithanincreaseintheopennessofanOlympichost. Thegamesdonotseemtoactas
simpleexportpromotion,butareinsteadassociatedwithanincreaseintwowaytradebetweenthe
hostandtherestoftheworld.
Nextwechangetheempiricalmodelbystrippingdownthegravitymodelbyincludingonly
bilateraldistance,importerpopulation,andimporterincomeascontrols. However,ourkeyfindingof
strongpositive coefficientspersists. Thesameistruewhenweaddregionaldummyvariables,taken
fromtheregionalgroupingsfromtheWorldBank(sothattheyincludeonlydevelopingcountries). In
particular,weincludedummyvariablesfor:EastAsiaandthePacific;EuropeandCentralAsia;Latin
AmericanandtheCaribbean;theMiddleEastandNorthAfrica;SouthAsia;andSubSaharanAfrica. We
thensuccessivelydeleteobservationsfordevelopingcountriesfrom:LatinAmericaortheCaribbean;
Africa;Asia;ortheMiddleEast. However,ourresultsdonotseemtobesensitivelyonobservations
fromaparticularregionoftheworld.
Wecheckfurtherforsamplerobustnessbyselectivelydroppingdifferentsetsofobservations.
WefirstdropallobservationsforimportersthatareclassifiedasindustrialbytheIMF. Wethen
consideronlymiddleincomeexporters(usingthe2008WorldBankdefinitionofacountrywithreal
annualGNIpercapitabetween$976and$11,905),andthenonlymiddleincomeimporters. Wethen
dropsmallexporters(definingasmallcountryasonewithfewerthanamillionpeople),poorexporters
(thosewithrealGDPpercapitaoflessthan$1000perannum),smallandthenpoorimporters,andthe
intersectionofthesefoursets. Finallywecreateamorebalancedsamplebyrestrictingourselvestothe
47countrieswithrelativelycompletebilateraldata. Noneoftheserobustnesschecksshakesthe
confidencewehaveinourbasicresults;wefindanegativeandsignificanteffectofhosting,butonlyin
theabsenceoffixedeffects. Wethencheckthesensitivityofourresultswithrespecttotime. We
separatelydroplate(post2000)andearly(pre1960)observations;againthisdoesnotdestroyour
findings.
Wefinishupthisanalysiswithfivefurtherexperiments. First,weweightourregressionsbythe
logoftheproductofthecountrypairrealGDP. Second,wedropoutlierobservations,definedasthose
wheretheresidualisfar(morethan2.5standarddeviations)awayfromzero. Wethenredefineourkey
Olympicsregressorintwoways.First,weconstructavariablewhichis0ifneitherinorjhashostedthe
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
9/30
8
Olympicsatorbeforetimet,1ifeitheriorjhashosted(atorbeforet),and2ifbothiandjhavehosted.
Thelastvariabledeliversinsignificantresultsforhostingthesummergameswhenfixedeffectsare
included;otherwise,ourresultsareinsensitive. Finally,weusetheGlickTaylorhistoricaldataset,which
includesbilateraltrade(butnotexport)datastretchingbackto1870.Usingthislongdatasetallowsus
toincorporateobservationsforcountriesthathavehostedthegamesatanypointduringthemodern
Olympiadera,notjustfromtheshorterpostwarperiod. However,ourfindingsremainresilient;
countriesseemtohavepermanentlyhighertradeflowsafterhavinghostedthegames.
Weconcludethatourresultsdonotseemtoresultfromsomesmallsubsetofthedataset,and
arerobusttoreasonablechangesinthesampleandexactspecificationofourempiricalmodel.
CountriesthathavehostedtheSummerOlympicgamesseemtohavehighertradethanothers.
4.2 Endogeneity
Aquestionarisesimmediately:CanthechoiceofvenuefortheOlympicgamesbetreatedas
plausiblyexogenous? Perhapsonlycountriesthatareopentotradearechosentohostthegames? We
attempttoaddressthispointmoredirectlybelowwithamatchingmethodologythatcomparesOlympic
hosts(astreatments)toothercountries(controls),andwealsopursueaneventstudyofrelevance.
Still,wenowmakeseveralpoints.
First,theendogeneitycritique(alongwithmuchofouranalysis)isprimarilycrosssectional,
whiletheempiricalfindingofrelevancetousisfoundinthetimeseriesbehavioroftrade.
Second,ouranalysisshowsthatcountriesbecomemoreopenaftertheOlympics. However,
cities(notcountries)bidtohostthegames. Indeed,fourAmericancitiesviedforthe1948Olympics
(whichwenttoLondon);fiveAmericancitiesappliedin1952(againwithoutsuccess),andsixin1956
(alsounsuccessfully).
3
Third,theIOCprovidesdetailsbothonhowitawardsthegamesandwhyacityshouldbe
interested. ThereisalonglistoftechnicalcriteriawhichareevaluatedbyanIOCcommittee;fewof
thesecriteriaarecloselyassociatedwithtrade. Informally,thereisalsogeographicbalancing,andthe
IOCseemstoawardperseverance(anumberofcitieshaveappliedrepeatedly). Wealsonoteinpassing
thatthereisconsiderablerandomnessinherentintheprocess.
Fourth,ourdatadoesallowustoconsidertheissueofreversecausalitydirectly.Inparticular,
wecanstatisticallyexaminewhethermoreopencountriesaremorelikelytobidfor,orobtainhosting
rightstotheOlympics. WeconductprobitteststothateffectinTableA2intheappendix. Inadditionto
includingopenness,wecontrolforcountrysizeandpercapitaincome. Wefindthatopennessenters
insignificantlythroughout,suggestingthatreversecausalityisnotanissue.
WhydoestheIOCthinkthatcitiesshouldbeinterestedinhostingtheOlympics? Theystate:
`Apartfromthesportingevents,themainreasonforapplyingforcandidacyliesinthepossibilitiesfor
economicdevelopmentandtourisminherentinsuchanevent. Forthisreason,andalsogiventhehigh
infrastructurecosts,onlyrichcountrieshavethemeanstomakeagoodreturnonsuchalarge
investment.4
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
10/30
9
`Twomainreasonsseemtomotivatemostapplicantcities,namelyinternationalrecognitionandincreased
opportunitiesforinvigoratedurbanandregionaldevelopment. Indeed,thehostcityhopestotake
advantageoftheeventtomaximizeitsfacilitiesduetotheconsiderableincomegeneratedbytheGames,
andtogiveitselfanenhancedimagetoattractfuturevisitors,consumersandpotentialinvestors...
OrganisingtheOlympicGamesisafantasticadvertisingopportunityforthehostcity... Moreover,
organizingtheOlympicGamesisanopportunityforthehostcityandcountrytoshowtheworldtheirability
toundertakeandorganisesuccessfullysuchanimportantevent. Thispromotionalaspectisoften
motivatedbythepoliticiansofthehostcountry,therebyexplainingtheheavyinvolvementofnationalgovernmentsintheorganisationandfinancingofOlympicGames."
4.3 OtherInternational`Megaevents
WhiletheOlympicsarehighlyvisible,theyarenottheonlymegaevent. Doothereventsdeliver
similarresults? OneobviousalternativetoconsideristheWorldCup,theonlyseriouscompetitorwith
theOlympicsfortitleofmostimportantinternationalsportingevent.
LiketheOlympics,theFIFAWorldCupisheldeveryfouryears. Itbeganin1930inmuchthe
smallscalewayasdidthemodernOlympicsgame,withthirteen(mostlyLatinAmerican)countries
participatinginatournamentheldinUruguay. The1934and1938tournamentswerealsorelatively
smallandregional,beingheldinEuropewithlimitedparticipationbytheLatins. Theeventonlyreally
tookoffin1950(the1942and1946eventswerecanceledforWorldWarII). Thiswasdueinparttoa
conventionofalternatingtheeventbetweentheAmericasandEurope.5 Weconstructdummy
variablesforcountriesthathavehostedtheWorldCuppostwarinamanneranalogoustothoseforthe
Olympicgames,andaddthemtoourspecification. WereportourresultsinTable3a.
TheresultsofTable3aareintriguing. TheeffectofhostingtheOlympicgamesremainspositive
andstatisticallysignificantforallthreespecifications;thecoefficientscontinuetoaveragearound.2.
TheeffectsofhostingtheFIFAWorldCuparesimilarinsign,size,andstatisticalprecision. Indeed,wecanneverrejectthehypothesisofequaltradeeffectsofhostingtheOlympicsandtheWorldCup.
Whataboutinternationaleventsthatdonotinvolvesports? Afterall,internationalexpositions
andworld'sfairshaveanolderpedigreethanthemodernOlympics,stretchingbackatleasttothe1851
GreatExhibitioninLondon'sCrystalPalace. Thedeclineofinternationalcommunicationand
transportationcostshaslargelymadeworld'sfairsobsolete,andtheyhavedeclinedinimportanceand
numberthroughoutthepostwarperiod(Expo67inMontrealiswidelyconsideredtobethemost
successfulexpoeverheld). Still,Table3baddscomparablyconstructedbinaryvariablesfor(twenty)
postwarexposandworld'sfairstoourdefaultspecification. Inallspecifications,thetradeeffectof
hostingaworld'sfairorexpoispositiveandstatisticallysignificant,thoughitissmallerthantheOlympiceffectontrade.
Thereseemtobeanumberofwaysinwhichacountrycanenhanceitstradebyhostingan
internationalmegaevent. ThetradeexpandingeffectsofhostinganeventliketheOlympicsseemtobe
large,andtheyarebroadlycomparabletothoseassociatedwithhostingtheFIFAWorldCup. Holdingan
expoorworld'sfairalsoseemstohavemuchthesameeffect. Givenourinitialdoubtsconcerningthe
benefitsofhostingamegaevent,wenowdiveintotheissuemoredeeply.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
11/30
10
5. UnsuccessfulCandidates
ItisdifficultforustobelievethathostingtheOlympicsactuallyhassuchalargeeffectontrade,
letaloneonethatenhancestradepermanently. Yetoursensitivityanalysisshowsthatourresultsare
sturdy.
Wenowtakeaskepticallookatourresultsfromadifferentangle. Toestimatetheeffectsof
theOlympicsontradeintheprevioussection,ourstatisticalmodelcomparedhostswithnonhosts.
Thisseemsareasonablestrategy,sincenotallcountrieshavehostedthegames(sothereiscross
sectionalvariation),andthecountriesthathosttheOlympicsdosoatdifferentpointsoftime(sothere
istimeseriesvariation). However,itmaybethatourregressionresultsareimplicitlycomparingapples
withoranges;countriesarenotrandomlychosentohosttheOlympics. Onewaytogetatthisissueisto
comparethetradepatternsofhostcountrieswiththosethatbidunsuccessfullyforthegames. Werefer
tothelatterasunsuccessfulcandidatecountries;candidatesaretabulatedinAppendixTableA1. Our
implicitassumptionincomparingthetradeeffectsofOlympichostingandcandidacyisthatfailed
candidaciesformavalidquasiexperimentalcounterfactualcontrolgroupforOlympichosts(afterthe
inclusionofotherconditioningvariables).
InTable4a,wereportresultswhenweaddasetofbinaryvariablesforcountriesthatwere
unsuccessfulcandidatestohosttheOlympics. Thesehavebeenconstructedinthesamewayasour
hostdummyvariables. Forinstance,Londonwasawardedthe1948summergames,sothesummer
hostvariableisunityforallBritish(export)observationsfrom1948throughtheendofthesample.
SinceLausannewasanunsuccessfulcandidateforthe1948games,allSwissobservationsfrom1948
alsotakethevalueofunityforthecomparablesummercandidatevariable. Wetabulateseparate
exportestimatesfor:a)beinganunsuccessfulOlympiccandidate;andb)actuallyhostingthegames.Intriguingly,alltheeffectsthatis,boththehostandthecandidatecoefficients aresignificantly
positive. Indeed,asthetestsatthebottomofthepanelshow,theyaresimilarinsize;thehypothesis
thathostandcandidateeffectsareequalcannotberejectedatthe1%confidencelevel(thoughitcan
beatthe5%levelfortwoofthethreecases,oncebecauseofalargerhostingeffect,andoncebecause
ofasmallerhostingeffect). Table4busesnarrowsthesetofunsuccessfulbeingcomparedtohosts;
onlytherunneruptotheactualhostchosenisused. Theresultsremainquiteconsistentwiththeidea
thatthehostingeffectontradeissimilarinsizetothatofbeinganotherwisesimilarbutunsuccessful
candidatefortheSummerOlympics.
5.1
Timing
of
the
Olympic
Effect
and
Exports
ThestatisticalanalysisofTable4indicatesthatbiddingtohosttheOlympicsseemstohavean
exporteffectcomparableinmagnitudetothatofhostingthegames. Weexplorethisresultfurtherin
Figure1,whichportraystheexporteffectsofbiddingfortheSummerOlympics,eithersuccessfullyor
unsuccessfully,inaneventstudy. InsteadofusingasingledummyvariabletoshowthepostOlympic
hostingeffectontrade(asinTables14above),weuseanunrestricteddistributedlagmodel,sothat
thelaggedimpactofbiddingforthegamesisallowedtovarydependingonthenumberofyearsbefore
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
12/30
11
orsincethebid. Sincebidstohostthegamesaretypicallysubmittedsevenyearsbeforetheactual
games,wemarkthesubmissionofbidswithaverticalline,andconsiderobservationsuptotwenty
yearsbeforeandfortyyearsaftertheactualgames. Wecontrolforothertradeeffectsthroughthe
gravitymodel(employedabove),portrayresultsforthreedifferentsetsoffixedeffects,andshowthe
coefficientsinthedistributedlagmodelalongwitha+/ twostandarderrorconfidenceinterval.
Exportsseeminsignificantlydifferentforbiddersandnonbiddersupuntilthetimethetimeof
theOlympicbid. Butafterthebidandbeforetheactualgamesexportsseemtorisesignificantly.
ExportsforOlympicbiddersremainsignificantlyhigherfordecadesafterthegames;theexactsize
dependsonthespecificationofthefixedeffects. Thiseffectdoesnotstemsimplyfromhosting;Figure2
isananaloguetoFigure1,butexaminesonlyunsuccessfulbidsforthegames. Thesimilaritybetween
thetwofiguresisstriking;inbothcases,theexporteffectstartsaroundthetimeofthebid,risesbefore
theactualgames,andappearstobeahighlypersistenteffect.
Thisisanintriguingresult,onewhichwefindreasonablyconsistentlythroughoutour
investigation. Itimpliesthatthe(sizeable)effectontradeseemstocomenotfromactuallyhostingthe
gamesbutfrombeingacountrythatbidsforthem. Moregenerally,signalingthatthecountryis
capableandwillingtohosttheOlympicsthroughahighlyvisibleinternationalbidforamegaevent
seemstobeassociatedwithasizeabletradeexpandingeffect. Indeed,theeffectofsendingthissignal
seemsbroadlycomparableinsizetoactuallyhostingthegames. Thisisconsistentwithourviewthat
anydirecttradeeffectofhostingthegamesissmall. Moreover,itsuggeststhattheobservedeffectsof
hostingamegaeventdonotseemtostemfromabigpush"typeofprocess[e.g.Murphyet.al.(1989).
Wedismissasimplausibletheideathateconomicfundamentals(e.g.,transportinfrastructure)relevant
fortradehavebeenimprovedbybothhostsandunsuccessfulcandidatesalikeduringtherunuptoa
seriousOlympicbid.6
6. FurtherRobustnessChecks
Inthissection,wemovebeyondtheconventionalgravitymodelandbrieflysubjectourresults
tothreefurthersensitivitytests. Givenourinterestinestimatingandcomparingthetradeeffectsof
bothhostingandbiddingforamegaeventliketheOlympics,wehavethreeparticularconcerns;a)
econometricissuesassociatedwiththegravitymodel;b)econometricissuesassociatedwithselection
biasandendogeneity;andc)thepresenceofourfindinginmultilateraldata.7
6.1 Tetradicestimates
Webeginbydealingwiththeproblemthatgravitymodelslikeoursmaybemisspecified
becauseofmonadic"problems. Theserefertoomittedfactorsthatarespecifictoasinglecountrybut
mayvaryovertime,suchasthoseassociatedwithmultilateralresistance"totrade,e.g.Andersonand
vanWincoop(2003). Todealwiththeproblem,weadoptthemethodoftetrads"advocatedbyHead
etal(2010).Withthismethod,consistentestimatorsofthecoefficientsofinterestcanbeattainedinthe
presenceofmultilateralresistancebycomparingexportobservationstoexportsforapairofbase
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
13/30
12
countriesforthesameyear(thetechniqueistetradicsinceonecomparestradeflowsforfour
countries). Thismethodavoidsthelargenumberofcoefficientestimatesthatwouldberequiredto
estimatethemonadiceffectsusingamoreconventionalfixedeffectsmethod,butdoesnoteither
infeasiblecomputationaldemandsornonlinearestimation;Headet.al(2010)providemoredetails.
Oneissuethatarisesintetradicgravityspecificationsisthedesignationofthebasecountries.Toensurethatourresultsarerobust,weusethreedifferentpairsofbasecountries:a)theUnitedStates
andtheUnitedKingdom;b)JapanandFrance;andc)GermanyandCanada. Asecondissueisthatthe
errortermsinourtetradsarelikelytobecorrelated,aserrortermsforindividualcountriesappear
repeatedlyacrossobservations.WethereforeusethemethodologyofHeadet.al(2010)tocorrectour
standarderrorestimates. Finally,thisestimationtechniquerequiresvariationacrossbothdyadsand
time,sothatthedummyvariableweusedforTable1isinadmissible;wesubstituteinsteadadummy
whichis0ifneitherinorjhasbidfortheOlympicsatorbeforetimet,and1ifeitheriorjhas.
OurresultsarereportedinTable5.Itcanbeseenthatwecontinuetoobtainpositiveand
statisticallysignificanteffectsofeitherhostingorcandidacyfortheOlympics,regardlessofourchoiceof
basecountries;ourpointestimateare,ifanything,implausiblylarge.
6.2 Matching:aTreatmentMethodology
Wenextuseatreatmentmethodology,comparingexportsforeithertreatedhostsor
candidatecountriestoexportsfortheirmatchedcontrolcounterparts. Candidateandhostcountries
fortheOlympicgamesarenotrandomlyselectedfromoursample;insteadtheychoosewhetherto
submitabidornot. Thusthereisthepossibilitythatourresultslargelyreflectselectionbias;usinga
matchingtechniqueallowsustohandlethisproblemmoreeffectively. Wematchobservationsusinga
stratificationtechnique. Ourvariablesusedformatchingcountrypairyearobservationsincludethe
logsof:distance,exporterandimporterpopulations,exporterandimporterrealGDPspercapita;anddummyvariablesforsharingacommonlanguageorborder.
Wedotwokindsofmatching:a)wematchactualOlympichostcountriestocandidates;andb)
wematchtheunionofhostsandcandidatestononcandidates. Ifhostandcandidatecountries
experiencesimilartradeboosts,weexpectthefirstexercisetoleadtosmalldifferencesinexports. If
biddingforamegaeventiskey,weexpectthesecondexercisetoresultinlargetradeeffects.
Ourresults,alongwithbootstrappedstandarderrorestimatesareshowninTable6. Wefind
thathostsoftheSummerOlympicsexperienceasmallincreaseinexportscomparedwithcandidatesfor
thegames.Thiseffectisstatisticallysignificantata5%(butnot1%)confidencelevel. Theseweak
resultscontrastwiththosewhichcomparehostsandcandidatestononcandidatecountries. Bidders
experienceaconsiderabletradeboostcomparedwithnoncandidates. Theseeffectsareeconomically
large(exportsrisebyabout20%)andaresignificantatthe1%confidencelevel.8
6.3 Multilateraldata
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
14/30
13
Asafinalrobustnesscheck,weexaminetheeffectofhostingorbeingacandidateforthe
Olympicsontheaggregateexport/GDPratio;wealsoconsiderbeingaWorldCuphost. Wedothisby
simplyregressingmultilateraldata(sothatanobservationisforaparticularcountryandyear)onthe
Olympicdummies.Toaccountforotherdeterminantsoftradeasafractionofoutput,wealsocontrol
forthelogsofpopulation,realGDPpercapita,andincludecountryfixedeffects(yeareffectsareadded
assensitivityanalysisinanothercolumn).
OurresultsareshowninTable7. WefindthatcountrieswhichhavehostedtheOlympicshave
tradeapproximately15%(=exp(.14)1)higher,otherthingsbeingequal. Weobtainasimilarbutslightly
largeeffectforunsuccessfulcandidacy;thiseffect(unlikethehostingeffect)issignificantlyabovezero
atconventionalconfidencelevels. Allinall,ouraggregateresultsmirrorthoseaboveinthesensethat
wecannotrejectthehypothesisthatthecoefficientestimatesonhostingandcandidacyarepositiveand
equivalent. Finally,wealsoobtainsimilarpositiveandstatisticallysignificantcoefficientestimatesfor
havinghostedtheWorldCupinallofourreportedregressions.
7.ASignalingModel
7.1Setup
OurresultssuggestthattheOlympiceffectontradeinthedataisnotassociatedwithhosting
thegamesbutratherwithbiddingforthem.Inthissection,wedevelopapoliticaleconomysignaling
modelconsistentwiththisresult.Ourmodelisofthe``burningmoney"type. Weassumethatcountries
thatintendtopursueliberaltradepoliciesinthefuturecansignalthisintentbyengaginginthecostly
activityofbiddingtohosttheOlympicGames.Thepayoffforsendingthissignalisthatcountrieswhich
expecttoliberalizereceiveincreasedinvestmentintheexportsector(thesectorwhosepricesareraisedbyliberalization).Underappropriateparameterconditions,weobtainaseparatingequilibriumwhere
countriesthatchoosetoliberalizealsochoosetobidfortheOlympics;thosethatprefertoremain
closedneithersendthesignalnordotheyliberalize.
Weintroduceatwosectorspecificfactorsmodelofasmallopeneconomyinwhich
liberalizationincreasespricesintheexportsectorandlowersthemintheimportcompetingsector.
Nationalgovernmentsdifferinthedegreetowhichtheyvaluegainstotheexportingsector,andwe
assumethattheycannotcrediblyrevealthesevaluationstopotentialinvestors.Thegovernmentineach
countrymakesadiscreteliberalizationandsignalingdecisionbasedonitsexpectationsconcerningthe
impactofliberalizationonitsutility.
Bothsectorsoftheeconomyproduceusingafixeddomesticfactor,whichcanbeconsidered
sectorspecificcapital.Puttycapital, k,ismobileacrosssectorsandearnsaninternationalmarketrate
ofreturn,whichisfixedat *r .9 Realoutputlevelsintheexportandimportcompetingsectorssatisfy
jy k ,where ' 0jy and " 0jy , , j x m .Forsimplicity,weassumethatallputtycapitalis
importedbydomesticentrepreneurs,whohaveclaimsonthefixedfactorsandearnanyresidualprofits
fromoperations.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
15/30
14
Asthecountryissmall,ittakesworldpricesasgiven.However,domesticpricesareafunctionof
thegovernment'sliberalizationdecision.Liberalizationraisespricesintheexportsectorandlowers
themintheimportcompetingsector. Priceswhilethenationisclosedtoexternalmarketsaredenotedc
xp andc
mp ,whileafteropeningtheyareo
xp ando
mp respectively,whereo c
x xp p ando c
m mp p .
Thetimingofthemodelisasfollows:First,thegovernmentdecideswhetherornottosubmita
bidtohosttheOlympics. Next,theprivateagentsmaketheirinvestmentdecisions,basedontheir
expectationsofthegovernment'sliberalizationdecision. Finally,thegovernmentmakesits
liberalizationdecision,thewinningOlympicshostisnamed,andthepayoffsaredetermined.
Toensuresubgameperfection,wesolvethemodelbackwards. Subsequenttoreceivingthe
government'ssignal,foreignputtycapitalisinvestedineachsectortoequatethevalueofmarginal
productinthatsectortotheworldinterestrate,suchthat,
* *( ) = (1 ).
l l
j j j p y k r (2)
where*; , , ,
l
jk j x m l c o representstheequilibriumamountofputtycapitalallocatedtosector j .
Since*/ '/ " 0; ,
j j jdk dp y p y j x m itfollowsthat* *c o
x xk k and* *c o
m mk k .
Giventhatthegovernment'sliberalizationdecisionconfirmsinvestorexpectations,thereturnto
thedomesticentrepreneursineachsector,l
j ,satisfies
* * *1 .l l j j j j j p y k r k (3)
Itiseasytoshowthatthereturnintheexport(importcompeting)sectorisgreater(lower)under
liberalization.10
Weassumethatthereputationcostofbeingawardedthegamesandbackingoutisprohibitive.
Assucheachnationexpectstohostthegamesconditionalonbeingawardedthem.11
Let h representthecostofhostingtheOlympics,netofanynonpecuniarybenefitofhosting
thegames,suchasnationalpride. Let representtheprobabilitythatacandidatewillbeawardedthe
righttohostthegamesconditionalonhavingsentthesignal.Theexpectedcostofsendingthesignalis
thereforeequalto h . Thegovernmentfinancesthecostofsendingthesignalbyimposingalumpsum
taxoneachsector,wheretheexportsectorpaysashare ofthecostofsendingthesignal, h ,and
theimportcompetingsectorpaystherest, 1 h .
Thegovernmentisassumedtohaveautilityfunctionthatisconcaveinearningsfromeach
sector.Thegovernment'sutilityfunctionisassumedtosatisfy
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
16/30
15
= ( ).g j j
j
U u v (4)
where 0, 0u u ,and , j x m .12Forsimplicity,wenormalizebysetting =1m ,anddefine
x asthemeasureofthedegreetowhichgovernmentutilityfavorstheexportsector.Weassume
thatthereisacontinuumofheterogeneouscountries [ , ] z z z ,whereahigherzindicatesahigher
valueofz
,thevalueofheldbythegovernmentofcountryz .Weassumethatz
issymmetrically
distributedontheinterval[ , ] withmeanvalue1.Fornotationalsimplicity,wedropthez
superscripts.
Priortosendingthesignal,gU thensatisfies
= ( ) ( ).c c
g x mU u v u v (5)
Aftersendingacrediblesignalandliberalizing,governmentutility, gU ,satisfies
= ( ) ( (1 ) ).o o
g x mU u v h u v h (6)
7.2 Equilibrium
Anequilibriumisasetofsignalandliberalizationdecisionsbyeachgovernmentthatmaximizes
itsexpectedutility,alongwithasetofinvestmentdecisionsinthetwosectorsconsistentwith
maximizingthereturnstothedomesticentrepreneurs,conditionalonthesignalofthegovernment.
Inanearlierversionofthispaper[RoseandSpiegel(2009)],wedemonstratethatthechangein
governmentutilitywithliberalizationismonotonicallyincreasingin.Define * asthevalueof z
whichleavesthegovernmentindifferentbetweenstayingclosedandnotsendingthesignal,and
sendingthesignalandliberalizing.Byequations(5)and(6),* satisfies
* ( ) ( (1 ) )=
( ) ( )
c o
m m
o c
x x
u v u v h
u v h u v
(7)
Wenextderivetheconditionsnecessarytoruleoutoffequilibriumpathoutcomes,wherethe
signalsentbythegovernmentfailstomatchitssubsequentliberalizationdecision.Let*
( )c c o
j j jv v k
( = , ) j e m representthevalueofrevenuesinsector j consistentwiththeoffequilibriumpathstrategy
ofsignalingliberalizationandthennotliberalizing,i.e.withcapitalconsistentwithliberalizationbut
pricesconsistentwithremainingclosed. Asshowninthelongerversionofthispaper,asufficient(but
notnecessary)conditiontoensurethatthisstrategyisnotchosenis
.c c
x xv v h (8)
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
17/30
16
Thisrestrictionrequiresthattheearningsoftheexportsectorwithoutliberalizationarehigherwhenthe
governmentdoesnotsendthecostlysignal.
Similarly,let*
( ); = ,o o c
j j jv v k j x m
represent i.e.thevalueofrevenuesinsector j consistent
withnotsignalingliberalizationandthenliberalizing.Asufficient(butnotnecessary)conditiontoensure
thatthisstrategyisnotchosenis
.o o
x xv h v (9)
Thisrestrictionrequiresthattheearningsoftheexportsectorwithliberalizationarehigherwhenthe
governmentsendsthe(costly)signal.Combined,theserestrictionsimplythatrevenuesintheexport
sectorarelowerunderbothoffequilibriumpathstrategies.
Giventheserestrictions,weobtainourfirstproposition:
Proposition1Thereexistsaseparatingequilibriumwherecountrieswith *z sendthesignalandliberalize,andcountrieswith *
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
18/30
17
liberalization;andthemarginalutilityoftheimportsectorafterliberalization.Theentiretermwillbe
negativeifandonlyif
* [ (1 ) ].
[ ]
o
m
o
x
u v h
u v h
(11)
Condition(11)requiresthat* (therelativevaluationofexportearningsheldbythe
governmentthatisjustindifferentbetweenliberalizingandnotliberalizing)islessthanorequaltothe
marginalrateofsubstitutionbetweenpostliberalizationearningsintheimportcompetingandexport
sectors.Intuitively,thisimpliesthatrelativeearningsintheimportcompetingsectoraresufficientlylow,
evenafteradjustingfortherelativeweightplacedonexportearnings(* ).Forexample,inthe
benchmarkcasewherethegovernmentvaluesearningsineachsectorequally*
( =1) ,thecondition
willbesatisfiedifearningsnetofexpectedsignalingcostssubsequenttoliberalizationarelowerinthe
exportsectorthanintheimportsector.
Giventhiscondition,weobtainoursecondproposition:
Proposition2Givenaseparatingequilibriumforallcountries [ , ] z z z ,andsatisfactionofcondition(11),anincreasein reduces * ,raisingthesetofcountriesthatchoosetosendthesignalandliberalize,whileif(11)isviolated,anincreasein increases * .
Theprooffollowsdirectlyfromequations(10)and(11).TheintuitionbehindProposition2liesin
thefactthatincreasesin improvethealignmentbetweenthecostsandbenefitsfromliberalization.
Thismayfavortheuseofbiddingfortheolympicsasasignalofopennessintentions.Thecosts
ofhostingtheGamesaretraditionallybornebythehostcity(usuallythecapital)inconjunctionwiththe
centralgovernmentofthehostcountry.Policymakersfromthesegroupsarelikelytobenefitfrom
liberalization.Intermsofourmodel,``megaevents"liketheOlympicsmaybehigh signals.
Tosummarize:ourmodelsuggeststhatcountrieschoosetobidforamegaeventinorderto
signalinvestorsabouttheirfutureliberalizationintentions.Undercertainparameterconditions,
governmentsthatwishtoliberalizecanprofitfromsendingthecostlysignalofbiddingtohostthe
games,whilethosethatdonotwishtoliberalizedonot;aseparatingequilibrium.Moreover,the
conditionsimplythattheprobabilityweightedcostofholdingtheOlympicsmustbesufficientlylargeto
dissuadegovernmentsthatdonotwishtoliberalizefromsendingafalsesignal.Thismotivatesthe
choiceofacostlymegaeventasasignalofliberalizationintentions.
Further,themodelalsodemonstratesthatdistributionalimplicationsmatter;theincidenceof
thecostofthesignalhasanimpactonitsdesirability.Thesignalmustbesufficientlycostlytotheexportsector,thesectorthatwouldbenefitfromthepolicychange,notjustcostlytothenation.Also,ifthe
governmentdoesnotfavortheexportsectortoogreatly,andliberalizationhassufficientdistributional
consequences,anincreaseinshareofexpectedcostofhostinggamesbornebytheexportsector
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
19/30
18
increasesthemarginalgovernment'swillingnesstobid.Thatis,thehigheristheexpectedburdenof
hostingtheOlympicsontheexportablesector,themoreattractiveisbiddingforamegaeventasa
signalofliberalization.
8. Conclusion
InJuly2001,BeijingwasawardedtherighttohosttheGamesoftheXXIXOlympiad. Justtwo
monthslater,ChinasuccessfullyconcludednegotiationswiththeWorldTradeOrganization,thus
formalizingitscommitmenttotradeliberalization. Noristhisaonceoffcoincidence. Romewas
awardedthe1960gamesin1955,thesameyearItalystartedtomovetowardscurrencyconvertibility,
joinedtheUN,and,mostimportantly,begantheMessinanegotiationsthatleadtwoyearslatertothe
TreatyofRomeandthecreationoftheEuropeanEconomicCommunity. TheTokyogamesof1964
coincidedwithJapaneseentryintotheIMFandtheOECD. Barcelonawasawardedthe1992gamesin
1986,thesameyearSpainjoinedtheEEC;thedecisiontoawardKoreathe1988gamescoincidedwith
Korea'spoliticalliberalization. ThecorrelationextendsbeyondtheOlympics;the1986WorldCupwas
heldinMexicocoincidentwithitstradeliberalizationandentryintotheGATT.13 Inthispaper,wemodel
thislinkagebetweenmegaeventsandliberalizationboththeoreticallyandempirically.
ThemotivationforhostingamegaeventliketheOlympicsseemselusivetoeconomists.
Plausiblymeasuredneteconomicbenefitsarerarelylargeandtypicallynegative;claimsofnon
economicbenefitsaredifficulttoverify. Yetinpracticecountriescompetefiercelyfortherighttohost
suchevents. Why? Thispaperidentifiesonepotentialexplanation;countriesthathostthegamesenjoy
asubstantivepermanentincreaseintrade the``OlympicEffect." Similarincreasesinopennessare
observedforcountriesthathostothermegaevents,suchastheWorldCupand,untilrecently,World's
Fairs. Foracountrypursuingatradeorienteddevelopmentstrategy,suchanoutcomewouldclearlybeattractive.
OurempiricalresultsshowthattheOlympiceffectisrobust;hostingthegamestendsto
increaseacountry'sopennesssubstantivelyandpermanently. Butwenotbelievethathostingthe
Olympicshasanactualcausaleffectontrade;whilehostingthegamesissufficienttoboosttrade,itis
notnecessary. Inpractice,wefindthatcountriesthatbidfortheOlympicsunsuccessfullyalso
experienceaboostintrade,comparabletothatreceivedbyactualOlympichosts. Thisfindingimplies
thattheOlympicEffectontradedoesnotstemfromachangeineconomicfundamentals,causedbythe
activityorinfrastructureassociatedwithhostingtheOlympics. Instead,ourempiricalfindingssuggest
thatbiddingfortheOlympicsisacostlypolicysignalthatisfollowedbyfutureliberalization. Weexplorethisconjectureinapoliticaleconomymodel,wheregovernmentschoosewhetherornottosignalfuture
liberalizationbyhostingamegaeventliketheOlympics. Wederivetheconditionsforaseparating
equilibrium,whereonlycountriesthatvalueliberalizationchoosetosendthesignalandliberalize. Our
modelalsosuggeststhatthesizeanddistributionalconsequencesofthistypeofsignalmayinfluenceits
desirability.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
20/30
19
Weclosewithanumberofcautions. First,ourmodelmakesnoclearstatementonthemeritof
publicsupportforhostingmegaevents. Second,thereareothermotivationsforhostingmegaevents
whichwehavenotmodeled;forinstance,ourtheorycannoteasilyexplainthebehaviorofcountries
thatsubmitrepeatedormultiplebidsforlargesportingevents,orwhyopeneconomiesbidformega
events. Finally,othersignalsofandroutestoliberalizationexist,andouranalysisdoesnotexaminethe
relativeeffectivenessofthesepaths;weleavesuchissuestofutureresearch.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
21/30
20
Table1:PermanentEffectofOlympicsonLogExportsinGravityModelSummerOlympicsHost .33**
(.04)
.24**
(.03)
.30**
(.04)
.19**
(.04)
.16**
(.04)
.34**
(.03)
.35**
(.04)
LogDistance 1.12**
(.02)
n/a 1.33**
(.02)
1.26**
(.02)
1.31**
(.02)
1.14**
(.02)
1.32**
(.02)
LogExpPopulation 1.07**
(.01)
.19**
(.06)
.25**
(.06)
.80**
(.02)
.78**
(.02)
1.08**
(.01)
.41**
(.06)LogImpPopulation .88**
(.01)
.80**
(.05)
.45**
(.05)
.91**
(.01)
.28**
(.05)
.79**
(.02)
.78**
(.02)
LogExpRealGDPp/c 1.55**
(.01)
1.24**
(.03)
1.25**
(.04)
1.16**
(.02)
1.20**
(.02)
1.60**
(.01)
1.27**
(.03)
LogImpRealGDPp/c 1.18**
(.01)
.87**
(.03)
.84**
(.03)
1.22**
(.01)
.85**
(.03)
.94**
(.02)
.95**
(.02)
CurrencyUnion 1.03**
(.10)
.56**
(.09)
.67**
(.10)
.82**
(.10)
.62**
(.09)
.80**
(.10)
.60**
(.09)
CommonLanguage .45**
(.04)
n/a .34**
(.03)
.44**
(.03)
.35**
(.03)
.39**
(.03)
.33**
(.03)
RTA .27**
(.03)
.29**
(.02)
.43**
(.03)
.33*
(.03)
.46**
(.03)
.42**
(.03)
.47**
(.03)CommonBorder .68**
(.08)
n/a .46**
(.08)
.45**
(.08)
.46**
(.08)
.70**
(.08)
.44**
(.08)
#Islands .17**
(.03)
n/a 3.86**
(.32)
.39**
(.03)
.55**
(.08)
.04
(.04)
.68**
(.08)
LogProductArea .07**
(.01)
n/a .57**
(.03)
.06**
(.01)
.02
(.02)
.07**
(.01)
.02
(.02)
CommonColonizer .58**
(.06)
n/a .75**
(.05)
.58**
(.05)
.77**
(.05)
.74**
(.05)
.74**
(.05)
CurrentlyColony .61*
(.24)
.38*
(.19)
.95**
(.25)
.75**
(.22)
.94**
(.22)
.81**
(.24)
1.02**
(.24)
EverColony 1.45**
(.10)
n/a 1.42**
(.09)
1.49**
(.09)
1.40**
(.09)
1.31**
(.09)
1.43**
(.09)CommonCountry .09
(.71)
n/a .95*
(.41)
.41
(.38)
.93**
(.35)
.35
(.62)
.99*
(.42)
YearEffects Yes Yes Yes Yes Yes Yes Yes
DyadicFixedEffects Yes
ExporterFixedEffects Yes Yes
ImporterFixedEffects Yes Yes
Exporter*TimeFixedEffects Yes Yes
Importer*TimeFixedEffects Yes Yes
R2
.61 .85 .69 .66 .69 .64 .69
RMSE 2.18 1.40 1.94 2.04 1.94 2.09 1.94
449,220bilateralannualobservationscovering196countries,19502006. Robuststandarderrors(clusteredby
dyads)inparentheses. Coefficientssignificantdifferentfrom0at.05(.01)markedwithone(two)asterisk(s).
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
22/30
21
Table2:SensitivityAnalysisofPermanentOlympicsExportEffectFixedEffects: Years Years,Dyads Years,Exporters,Importers
SubstituteImportsfor
Exports
.62**
(.04)
.50**
(.04)
.65**
(.05)
StrippedDownGravityModel .30**
(.04)
.22**
(.04)
.27**
(.04)
AddRegionalDummies .21**(.04) .24**(.03) .30**(.04)
DropLatinAmerican,
CaribbeanImporters
.28**
(.04)
.22**
(.04)
.29**
(.04)
DropAfricanImporters .37**
(.04)
.26**
(.04)
.34**
(.04)
DropAsianImporters .35**
(.04)
.26**
(.04)
.33**
(.04)
DropMiddleEastern
importers
.33**
(.04)
.24**
(.04)
.30**
(.04)
DropIndustrialImporters .31**
(.04)
.25**
(.04)
.27**
(.04)
OnlyMiddleIncome
Exporters
.39**
(.13)
.88**
(.14)
.86**
(.14)
OnlyMiddleIncomeImporters
.33**(.05)
.28**(.05)
.27**(.06)
DropSmall(
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
23/30
22
Table3:ThePermanentEffectsofOtherMegaEventsonLogExports
A. EffectsofHostingOlympicsandWorldCupFixedEffects: Years Years,Dyads Years,Exporters,
Importers
OlympicEffect .30**
(.04)
.19**
(.03)
.22**
(.04)WorldCupEffect .31**
(.03)
.18**
(.03)
.27**
(.03)
Olympic=WorldCup
Effect?(pvalue)
.80 .95 .39
B. EffectsofHostingOlympicsandWorldsFair/ExpoFixedEffects: Years Years,Dyads Years,Exporters,
Importers
OlympicEffect .26**
(.04)
.21**
(.03)
.26**
(.04)
WorldsFair/ExpoEffect .19**(.04)
.07**(.03)
.09**(.03)
Olympic=Worlds
Fair/ExpoEffect?(pvalue)
.24 .00** .00**
Coefficientonpermanenteffectofhavinghostedamegaevent. Standarddatasetincludesupto449,220bilateral
annualobservationscovering196countries,19502006. Robuststandarderrors(clusteredbydyads)in
parentheses. Coefficientssignificantdifferentfrom0at.05(.01)markedwithone(two)asterisk(s).
Regressandislogrealbilateralexports. Controlsincludedbutnotrecorded:LogDistance;LogExporterPopulation;
LogImporterPopulation;LogExporterRealGDPp/c;LogImporterRealGDPp/c;CurrencyUniondummy;Common
Languagedummy;RTAdummy;CommonLandBorderdummy;#Islands;LogProductArea;CommonColonizer
dummy;CurrentColonydummy;EverColonydummy;andCommonCountrydummy.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
24/30
23
Table4a:ThePermanentEffectsofOlympicHostingandCandidacyonLogExportsFixedEffects: Years Years,Dyads Years,Exporters,
Importers
OlympicHostingEffect .30**
(.04)
.20**
(.03)
.24**
(.04)
OlympicCandidacyEffect .20**(.03)
.25**(.03)
.34**(.03)
Hosting=CandidacyEffect?
(pvalue)
.02* .26 .04*
Table4b:EffectofHostingandbeingRunnerUponLogExportsOlympicHostingEffect .33**
(.04)
.23**
(.03)
.29**
(.04)
RunnerUpEffect .09*
(.04)
.22**
(.04)
.31**
(.04)
Hosting=RunnerUp
Effect?(pvalue)
.00** .75 .68
Coefficientonpermanenteffectofhavinghosted/beenacandidate/beentheleadingrunnerupfortheSummerOlympics. Standarddatasetincludesupto449,220bilateralannualobservationscovering196countries,1950
2006. Robuststandarderrors(clusteredbydyads)inparentheses. Coefficientssignificantdifferentfrom0at.05
(.01)markedwithone(two)asterisk(s).
Regressandislogrealbilateralexports. Controlsincludedbutnotrecorded:LogDistance;LogExporterPopulation;
LogImporterPopulation;LogExporterRealGDPp/c;LogImporterRealGDPp/c;CurrencyUniondummy;Common
Languagedummy;RTAdummy;CommonLandBorderdummy;#Islands;LogProductArea;CommonColonizer
dummy;CurrentColonydummy;EverColonydummy;andCommonCountrydummy.
Table5:TetradicPermanentEffectsofOlympicHosting/CandidacyonLogExportsBaseExporter USA Japan Germany
BaseImporter
UK
France
Canada
Effectof
Host/Candidacy
.44**
(.04)
.45**
(.04)
.76**
(.04)
Observations 534,820 523,406 514.703
Bilateraldatasetcovers196countries,19502006. Robuststandarderrorsinparentheses. Coefficientssignificant
differentfrom0at.05(.01)markedwithone(two)asterisk(s).
Regressorsincludedbutnotrecorded:CurrencyUniondummy;andRegionalTradeAgreementdummy. Year
effectsincludedbutnotrecorded.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
25/30
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
26/30
25
Figure1:TheOlympicEffectonExports
0
.5
1
-20 Bid Games 20 40
No Fixed Effects
0
.3
.6
-20 Bid Games 20 40
Country Fixed Effects
0
.3
.6
-20 Bid Games 20 40
Country-Pair Fixed Effects
449,200 bilateral observations (1950-2006, 196 countries), gravity controls
Mean, with +/-2 se Confidence Interval
Export Effect of Any Summer Olympic Bid
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
27/30
26
Figure2:UnsuccessfulOlympicBidsandExports
0
.5
1
-20 Bid Games 20 40
No Fixed Effects
0
.3
.6
-20 Bid Games 20 40
Country Fixed Effects
0
.3
.6
-20 Bid Games 20 40
Country-Pair Fixed Effects
449,200 bilateral observations (1950-2006, 196 countries), gravity controls
Mean, with +/-2 se Confidence Interval
Failed Summer Olympic Bids and Exports
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
28/30
27
AppendixTableA1:HostsandUnsuccessfulCandidateCitiesforPostWarSummerOlympicsGames
Host UnsuccessfulCandidates
1948 London,UK Baltimore,Lausanne,LosAngeles,Minneapolis,Philadelphia.
1952 Helsinki,Finland Amsterdam,Chicago,Detroit,LosAngeles,Minneapolis,Philadelphia.
1956 Melbourne,
Australia
BuenosAires,Chicago,Detroit,LosAngeles,MexicoCity,Minneapolis,Philadelphia,
SanFrancisco
1960
Rome,Italy Brussels,Budapest,Detroit,Lausanne,MexicoCity,Tokyo.
1964 Tokyo,Japan Brussels,Detroit,Vienna.
1968 MexicoCity,Mexico BuenosAires,Detroit,Lyon.
1972 Munich,Germany Detroit,Madrid,Montreal.
1976 Montreal,Canada LosAngeles, Moscow.
1980 Moscow,USSR LosAngeles.
1984 LosAngeles,USA None
1988 Seoul,Korea Nagoya.
1992 Barcelona,Spain Amsterdam,Belgrade,Birmingham,Brisbane,Paris.
1996 Atlanta,USA Athens,Belgrade,Manchester,Melbourne,Toronto.
2000 Sydney,Australia Beijing,Berlin,Istanbul,Manchester.
2004 Athens,Greece BuenosAires,CapeTown,Rome,Stockholm.
2008
Beijing,China Istanbul,Osaka,Toronto,Paris
Dataavailableathttp://www.olympic.org/uk/games/past/index_uk.asp?OLGT=1&OLGY=1992
AppendixTableA2:DeterminantsofBiddingforandHostingtheSummerOlympics
Treatment(1) Control(0) Log(Export/GDP) Log(Population) Log(RealGDPp/c)
Host NonHost .43
(.31)
.24
(.13)
.68**
(.20)
Bidder(Hostor
UnsuccessfulCandidate)
NonBidder .31
(.20)
.42**
(.09)
1.01**
(.15)Host Unsuccessful
Candidate
.64
(.63)
.23
(.28)
.43
(.52)
Probitestimation;yeareffectsincludedbutnotrecorded. Coefficientswithstandarderrorsinparentheses;
coefficientssignificantlydifferentfromzeroat.05(.01)levelmarkedwithone(two)asterisk(s).
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
29/30
28
References
Alesina,A.,andA.Drazen(1991)"WhyareStabilizationsDelayed?"AmericanEconomicReview,81,11701188.
Anderson,J.E.,andE.VanWincoop(2003)GravitywithGravitas:ASolutiontothe
BorderPuzzle"AmericanEconomicReview,93(1),170192.Banerjee,A.(1992)ASimpleModelofHerdBehavior"TheQuarterlyJournalofEconomics,107(3),797817.
Bartolini,L.,andA.Drazen(1997)CapitalAccountLiberalizationasaSignal"AmericanEconomicReview,87(1),138154.Coates,D.(2007)StadiumsandArenas:EconomicDevelopmentorEconomicRedistribution?"
ContemporaryEconomicPolicy,25(4).Head,K.,T.Mayer,andJ.Ries(2010)TheErosionofColonialTradeLinkagesAfterIndependence
JournalofInternationalEconomicsforthcoming.Humphreys,J.M.,andM.K.Plummer(1995)TheEconomicImpactofhostingthe1996Summer
Olympics"availableathttp://www.selig.uga.edu/forecast/olympics/OLYMTEXT.HTM.
Krugman,P.(1998)TheConfidenceGame"TheNewRepublic.Lucas,R.E.J.(2009)TradeandtheDiffusionoftheIndustrialRevolution,"AmericanEconomicJournal:Macroeconomics,1(1),125.
Mukand,S.W.(2006)Globalizationandthe'ConfidenceGame'"JournalofInternationalEconomics,70,406427.
Murphy,K.M.,A.Shleifer,andR.W.Vishny(1989)IndustrializationandtheBigPush"JournalofPoliticalEconomy,97(5),10031026.Owen,J.G.(2005)EstimatingtheCostandBenefitofHostingOlympicGames"TheIndustrialGeographer,1(3),118.
Preuss,H.(2004)TheEconomicsofStagingtheOlympics(EdwardElgar,Northampton).Wacziarg,R.andK.H.Welch(2008)TradeLiberalizationandGrowthWorldBankEconomicReview22(2),187231.
8/7/2019 The Olympic Effect - Andrew K Rose & Mark M Spiegel -2010
30/30
Endnotes
1 Sinceweincludebothtime andcountry/dyadspecificfixedeffects,thiscanbeviewedasadifferencein
differencesestimator.
2 WehavealsoestimatetheSummerOlympiceffectalongwithaseparateWintereffectoracombinedSummer
andWintereffect;earlierversionsofthepapercontaintheresults. WhiletheWintereffectisgenerallyeconomicallyandstatisticallyinconsequential,thecombinedWinterandSummereffectisquitesimilartothe
effectassociatedwithonlytheSummergames.
3 Still,supportfromthenationalauthoritiesisnowaprerequisiteforaseriousapplication.
4 P1ofCandidateCitiesandVenuesfortheWinterOlympicsavailableat
http://multimedia.olympic.org/pdf/en\_report\_666.pdf
5 WenoteinpassingthattherehasbeenrelativelylittlecompetitiontohosttheWorldCup,sothatweareunable
toplausiblycomparehostsandunsuccessfulcandidates,aswedobelowfortheOlympics.
6
ThefundamentalsofrelevancefortradeseemessentiallyunrelatedtothecharacteristicsofagoodOlympicbid.7 Wefocusourattentionontheeffectsofmegaeventsontraderatherthantradepolicysincethelatterisdifficult
tomeasure. WehaveexperimentedwiththeWacziargWelch(2008)measureoftradeliberalization,andfindthat
itispositively(andoftensignificantly)correlatedwithpastOlympichostingorcandidacy,evenaftertakinginto
accountfactorslikecountrysizeandincometime andcountryeffectscontrollingfor. Thisresultisquite
consistentwiththemodelwedevelopbelow. However,wedonotconsiderthisavenueofresearchtobeworth
pursuinguntilwehavebetterempiricalmeasuresandmodelsoftradeliberalization.
8 Forcomputationalreasons,weestimateourmatching effectsusingonlyexportstodevelopedcountries.}
9 Weassumethatcapitalmarketsareopen.
10
Theinducedcapitalinflowsareonlyoneexampleofthepotentialbenefitsthatmightresultfromsendingacrediblesignaloftheapproachingliberalization.Thesignalmayalsoinduceadditionaldomesticinvestment,or
othertypesofcooperativebehaviorbydomesticagents.
11 BiddingfortheOlympicsandnothostingthemwouldbehighlyembarrassingandwouldadverselyimpacta
nation'sinternationalreputation.Theseareinfrequent,highlyvisibleeventswithlongleadtimes. Weknowof
onlyonecounterexample;Denverbackedoutofhostingthewintergamesin1976whenvotersrejectedabond
issuetofinancetheevent. Wealsoruleoutsignalsbycountriesthathavenochanceofwinningtherightstohost
thegamesas ,theprobabilityofwinninghostingrightsconditionalonbidding,iscommonacrosscountries.
12Thewelfarefunctionisspecifiedintermsofearningsineachsectorforsimplicity.Thisfunctionisassumedto
encompassallproceedsinthesesectors,includinganyredistributionofgovernmenttariffrevenuesandtheeffects
ofothertradedistortionsthatyielddifferentialvaluationsofdomesticpricesunderclosedandopenpolicies.
13 TherearealsoexamplesofunsuccessfulOlympiccandidateswholiberalized. Forinstance,SouthAfricabegana
dramatictradeliberalizationinthemid1990swhileitmounted(andlost)itsbidtohostthe2004Olympics(the
choiceofAthenswasannouncedinSeptember1997).