Page 1
IfYouBuildIt,WillSoccerFansCome?EvaluatingtheFactorsthatMakeMarketsViableforMajorLeagueSoccer
by
EricRoden
Abstract
MajorLeagueSoccer(MLS)isgrowingquicklybyimprovingitsexistingfranchises,andalsobyexpandingtonewmarkets,addingnewfranchises.MLShasthepotentialtoseesignificantgainsfromthisexpansionintonewmarkets;however,aswitheverybusinessdecision,therearerisks.Thereisalwaysthepotentialthatanewfranchisemayfailifplacedinthewrongcity.Thisisaconcernwithexpansioninanyprofessionalsportsleague,butparticularlysoinoneasyoungasMajorLeagueSoccer.Giventhesizablerisk,itissurprisingthatrelativelylittleresearchexistsontheevaluationofpotentialmarketsfornewsportsfranchises.Thisstudyfillsthisvoidbyanalyzinghowavarietyoffactorsthatcharacterizeageographicmarket–i.e.,population,communityattachment,thepresenceofotherprofessionalsportsfranchises,householdentertainmentspending,andeducationlevel(atametropolitanstatisticalarealevel)–affectMLSattendance.Resultsindicateapositiverelationshipbetweenpopulation,entertainmentspending,andeducationlevelwiththesuccessofMLSfranchiseswhileanegativerelationshipexistsbetweenthepresenceofotherprofessionalsportsfranchisesandcommunityattachmentwithfranchisesuccess.Subsequently,theobservedrelationshipsareusedtopredictwhichgeographicmarketswouldbemostsuitableforfutureMajorLeagueSoccerexpansioninAmerica,withAtlanta,SanFrancisco,Miami,andPhoenixappearingmostattractive.Keywords:MajorLeagueSoccer,newmarketentry,marketviability,professionalsportsexpansionSubmittedunderthefacultysupervisionofProfessorJoelWaldfogel,inpartialfulfillmentoftherequirementsfortheBachelorsofScienceinBusiness,summacumlaude,CarlsonSchoolofManagementUniversityofMinnesota,Spring2016.
Page 2
1
I.Introduction
SoccerisgrowinginAmerica;participationinthegamehasrisensteadilyinrecent
years.USYouthSoccerreportedanincreaseofover200,000registeredplayersfrom2013
to20141,andthetopprofessionalleagueinthecountry,MajorLeagueSoccer(MLS),has
seenitspergameaverageattendancegrowthelastthreeyearsrunning,increasingby
almost2,500lastyearalone.ThisrecentsurgehasindicatesthattheMLSisnowthefastest
growingAmericanprofessionalsport2.MLSattendancehasseenasteadyrisesinceits
foundingin1993(Figure1)andMLScommissionerDonGarberistryingtoensurethatthe
leaguecapturesasmuchoftheAmericanmarketaspossiblebycontinuingtoexpandthe
brandandimprovingprofitability.
Figure1:MLSAverageSingleGameAttendanceOverTime
MLSattendancehasseengradualgrowthoverthelastdecadefollowingasharpdropinattendanceinitssecondseason.Note:Attendancestartsat10,000,notat0.
1http://www.usyouthsoccer.org/media_kit/keystatistics/2https://en.wikipedia.org/wiki/Major_League_Soccer_attendance#MLS_attendance_vs._other_North_American_professional_sports_leagues
10,000
12,000
14,000
16,000
18,000
20,000
22,000
24,000
AverageperGameAttendanceOverTimeforMLS
Page 3
2
Sinceitsfirstseason,20yearsago,theMLShasrepeatedlyexpandedintonew
markets.Startingwithonly10teamsin1993,theleaguegrewto20teamsin2015.The
leagueplanstoexpandto24teamsby2018,withnewfranchisesalreadyannouncedfor
Atlanta,Minneapolis/St.Paul,andMiami.MLS’spathtothispointhasbeenbumpy,as
severalfranchiseshavefoldedoverthecourseoftheleague’shistory.Forexample,aftersix
yearsintheleagueandrelativesuccessonthefield,theTampaBayMutiny,afounding
memberoftheMLS,foldedin2001;theMiamiFusiondroppedfromtheleaguethatsame
year,afteronlyfouryearsofplay,thelastofwhichsawthemfinishinfirstplaceattheend
oftheregularseason;andin2014,LosAngelesbasedChivasUSAfolded,despitethe
comparativesuccessofcross-townrivalstheLAGalaxy,whomaintainoneofthelargestfan
basesinMLS.Althoughalackofonfieldsuccessmayhaveplayedapartinthedownfallof
someofthesefranchises,adrivingfactorwassimplelackofattendance.
Incontrast,successfulfranchises,suchasSeattleSoundersFC,SportingKansasCity,
andthePortlandTimbers,routinelypacktheirstadiumsandappearstableenoughtolast
yearsintothefuture,despitebeinglocatedinsmallermarkets.Thisinconsistencybegsthe
question:whatfactorsaremostimportantindeterminingtheviabilityofamarketfor
MajorLeagueSoccerfranchises,andwhatpotentialmarkets,presentlylackingteams,
wouldbemostviableforexpansion?
Inordertodeterminewhatfactorsmakeamarketviable,andultimatelywhich
marketswouldmakethebestexpansionlocations,itisperhapsusefultoexaminewhich
factorsassociatewithhighattendanceinexistingMLSmarkets.Moreover,ifwemakethe
assumptionthatamodelofattendanceincurrentMLSmetroareascouldbeextendedto
predictattendanceatnewsites,theresultsofthisanalysismightbeusedtoinformthe
Page 4
3
selectionoffranchiseexpansionlocationsinthefuture.ItislikelythatsomeofthebestMLS
marketsarealreadyplayinghosttoateam;however,theselectionprocessthusfarhas
beenmostlybasedonmetropolitanpopulationsizeandthepresenceofawilling
ownershipgroup.ComparingPanelAtoPanelBinFigure2shows,attendancedependson
acomplexarrayoffactors,beyondsimplemetropopulation.Youngfranchises,suchas
OrlandoCityFC,SeattleSounders,andPortlandTimbers,whorankamongthemost
attendedMLSfranchises,despiteallbeinginitiallyoverlooked,lendcredencetotheidea
thatgoodmarketslikelyremainuntapped.
Figure2:ComparingPopulationtoAttendanceinExistingMLSLocationsThetwofigurescomparethemetropolitanareapopulationofallcurrentUScitieswithaMLSfranchisetotheir2015MLSattendance.Comparingthetwomapsshowsthatalthoughtheremaybesomerelationshipbetweenpopulationandattendance,itisfarfromtheonlyfactoratplay.PanelA:2014MetropolitanStatisticalAreaPopulationsforMLSCities
Page 5
4
PanelB:2015AttendanceforMLSCities
Note:NewYorkCityhastwoteams;however,forreadability,onlytheattendanceforNewYorkCityFCisshown,thelargerofthetwoin2015.NYCFCaveraged29,016,whiletheNewYorkRedBullsaveraged19,657.GraphiccreatedusingTableaupublic.
Todevelopamodelofattendance,generaldemographicandconsumerinformation
forcitiesthatcurrentlyhostMLSteamswasanalyzed.Specifically,theimpactthat
communityattachment,theavailabilityofsubstituteentertainmentoptions(i.e.,other
sportsfranchises),educationlevel,population,andpsychologicalneedforaestheticshave
onMLSattendancewasinvestigated.Thisanalysiswasperformedviamultipleregression,
assessingtherelationshipbetweenfanattendanceandlocationfactors.Furtheranalysis
alsoexaminedtheabove-describedfactorsabilitytopredictwhichmarketshavea
franchise(i.e.thepresenceofanMLSfranchise).Thedevelopmentofthesecondmodel
allowsforcomparisonbetweenusingattendanceasadependentvariableandusingjust
thepresenceofateaminthemarket.Subsequentanalysisassessedwhichmarketsmake
thebestfitforfutureMLSexpansionwithconsiderationgiventobothmodels.
Page 6
5
Althoughthereexistsalargeamountofresearchthatstudiesthefactorsthat
motivateconsumerstoattendsportingevents,andasmallerbodyofresearchthathas
touchedonprofessionalsportsexpansion,thisstudyisnovelinthatitprovideswhatisto
myknowledgeafirstattempttoidentifyandcomparepossiblenewmarketsforthe
creationofsoccerfranchises,basedonanassessmentofthecommonfeaturesofcitiesthat
presentlyhostsuccessfulfranchises.Thefindingsofthisstudythereforehavethepotential
toinformthelocationchoicesfornewMLSfranchises.Thiswouldimpactlocalandnational
soccerenthusiastsbyprovidingimprovingtheirabilitytomakeacaseforreceivinganMLS
franchise.Moreover,inturn,itwouldbenefitthoselocaleconomies,throughthejobs
associatedwiththecreationofanynewfranchise.
Morespecifically,thisstudyaddressedfivehypothesesandutilizedtheresultsto
evaluatethepredictedsuccessofahypotheticalnewMLSfranchiseinvariouspotential
markets.Thefivehypothesesexaminedareasfollows:
• Hypothesis1:CommunityAttachment-Apositiverelationshipexists
betweencommunityattachment,asmeasuredbythepercentageofthe
populationwhowasborninthecityinwhichtheycurrentlylive,andMLS
franchisesuccess.
• Hypothesis2:TotalSports-Anegativerelationshipexistsbetweenthe
numberofotherprofessionalsportsteamsinamarketandMLSfranchise
success.
• Hypothesis3:NeedforAesthetics-Apositiverelationshipexistsbetweenthe
needforaestheticsandMLSfranchisesuccess.
Page 7
6
• Hypothesis4:Population-Apositiverelationshipexistsbetweenpopulation
sizeandMLSfranchisesuccess.
• Hypothesis5:Knowledge-Apositiverelationshipexistsbetweendesireto
acquireknowledgeandMLSfranchiseattendance.
ThetworegressionmodelsgeneratedsubsequentlybothsuggestthatAtlanta,San
Francisco,Phoenix,andMiamiholdthegreatestpotentialastopexpansionlocations.
II.LiteratureReview
TodeterminewhichfactorshavethegreatestimpactonattendanceinMLSmarkets
andinpredictingthepresenceofanMLSfranchise,theexistingliteratureonsports
consumptionandmotivationwasconsulted,resultinginseveralinsights.Mostofthe
existingliteraturefocusedontheimpactvariousfactorshaveoncertainpsychological
outcomes,suchasattachmentandmotivation,inaspecificpopulation.Additionally,the
psychologicalstudiestypicallyfocusedonasinglemarketandasinglesportsteam.Finally,
theliteraturearoundstudiesthatspecificallyfocusedonunderstandingprofessionalsports
expansionthrougheconomicmodelingfailedtoaccountforthedegreeofsuccessof
existingfranchises,insteadfocusingonjustthepresenceorlackofafranchiseinametro
area.
Thedesignofthisstudyaddressesbothoftheissuesmentionedabove.First,the
studydesignexaminesdatafrommultiplemarketsandmultipleteamstodevelopaclearer
pictureofthebreadthofapplicabilityofthepsychologicalfindings.Second,thisstudy
comparestheresultsofmodelsbasedonameasureofsuccess(attendance)tothosebased
onthepresenceofafranchise.
Page 8
7
i.Theonemarket,onesportapproach
Asmentionedabove,onerecurringtrendthroughtheliteraturewasthat
researcherstypicallylimitedtheirstudiestoonemarket,andthefansofonespecific
franchise.Forexample,James,KolbeandTrail(2002)studiedspecificallyindividual
supportersofaspecificnewMajorLeagueBaseballfranchise.KimandTrail(2010)usedan
evenmorespecificapproach,focusingtheirresearchonthefansinattendanceatasingle
WNBAgame.Funk,BeatonandAlexandris(2012)tookthebroadestapproachintheir
study,whichexaminedconsumersofvariousAustralianrugbyandfootballteams.This
paper,however,stilltreatedthesampleasonepopulationanddidnotlooktoexaminethe
differencesbetweenthetwosports'fansorthevariousmarketsstudied.Keepingthe
studiestoalimitednumberofmarketsandfranchiseslikelymadeiteasiertoidentify
factorsthatwereaffectingthepopulationbecauseitheldsocio-culturalandgeographic
factorsrelativelyconstant.Limitingthestudiesbreadthofmarketsandfranchisesand
focusingonthepsychologicaloutcomescomparedtoactualattendancealsohadits
weaknesses.Theliteratureleftagapindeterminingwhatfindingscanbegeneralizedto
otherpopulationsandalsodidnotallowforcross-marketcomparisons.
Additionally,theliteraturetypicallytookamorepsychologicalapproachtosports
consumptionbehaviorratherthandirectlyfocusingonattendance.UhlmanandTrail
(2012),forexample,focusedonvariables’correlationwithmotivationratherthanactual
attendanceandJamesetal.(2002)focusedprimarilyonlevelsofpsychological
involvement.KimandTrail(2010)focusedonthecausesofvariationinattendanceand
usedpsychologicalvariablesexaminedinotherresearch,suchasattachmenttotheteam,
Page 9
8
implyingthatincreasedmotivationultimatelyleadstoincreasedconsumption.
Additionally,focusingonmotivationandsimilarlyabstractconceptsprovidedabasisfor
determiningthefactorsthatwerelikelytobethemostsignificantindeterminingwhat
makesagoodMLSmarket.
ii.Similarfactorsinfluencesportsconsumptionacrossstudies
Anothertrendintheexistingresearchwastheemergenceofseveralfactorswhich
influencesportsconsumption,andthusattendance,inindependentstudieswith
independentpopulations.Forexample,Funketal(2010)andUhlmanandTrail(2012)
bothdeterminedthatthefans’senseofesteemandsuperiorityinfluencedconsumption.
UhlmanandTrailalsofoundthatconsumptioncorrelatedwithvariousformsof
attachment,consistentwithKimandTrail(2010).Theliteraturereviewuncoveredfurther
similaritieswhichinformedthehypothesesofthisresearch.Idiscussthosesimilaritiesin
coordinationwiththecorrespondinghypothesesbelow.Findingtheseconsistencies
allowedfortheevaluationofmarketsbasedonthesesharedfactors.
Overall,thesefindingsprovidedthebasisfortherestofthisresearch,buttheyalso
presentedsomeproblemsandillustratedareaswherethecurrentliteraturefellshort.With
theexceptionofKimandTrail(2010),themajorityofpriorresearchfocusedonfactors
thatarepositivelycorrelatedwithconsumption.Incontrast,KimandTraildiscussthe
constraintsthatlimitindividuals'abilitytoconsumesports,suchasdisposableincome.
Thesearealsoimportantfactorstoconsiderbecausetheyclearlyimpactattendance.
Additionally,factorsbeyondbasicdemographicconsiderationsimpacted
individual’smotivationtoconsumesports.Funketal(2012)andUhlmanandTrail(2012),
Page 10
9
foundthatfactorsrelatedtothefranchises’accomplishmentsalsoplayedanimportantpart
indeterminingconsumptionhabits–i.e.,thesuccessofthefranchisehadasignificant
impactonattendance.KimandTrail(2011)supportedthisbyfindingthatateam’ssuccess
actedasamotivatortoattendanceandon-fieldfailureactedasaconstraint,limiting
attendance.Infact,thestudyconductedbyKimandTrail,theyalsoevaluatedvarious
factorsimpactonattendanceandfoundthatlackofsuccessaccountedfor10%ofthe34%
ofvariancetheirmodelexplained.Theimpactrankedonlybehindteamattachmentintheir
study.Ateam'ssuccessisnotaninherentfeatureofamarketinwhichateamislocated,
butratherafunctionofmanagementandoperations.Therefore,itwasimportanttocontrol
forthisinordertocomparetheviabilityofmarkets,whichisindependentoftheimpactof
managementdecisionsoncethefranchiseenteredthemarket.Forexample,ifateamis
managedexceptionallywellandconsistentlyplacesnearthetopoftheleaguethenthey
wouldhaveahigherattendancethanifthesameteamwereperformingpoorly,andthat
variationhaslittletodowiththeviabilityofthemarketbutratherthemanagementofthe
franchise.
iii.Neglectingfranchisesuccesstofocusonjustthepresenceofafranchise
Thefinalinsightthatemergedfromtheliterature,andperhapsthemostimportant,
showedthatthefewstudiesthatfocusedonprofessionalsportexpansionusedthe
presenceofateamtoindicateagoodmarket,notthesuccessofthoseteams.Bruggink&
Zamparelli(1999),Baehr(2006),andDavis(2006)alltookthisapproachinmodeling
professionalbaseball,professionalsoccer,andminorleaguebaseballrespectively.Each
modelusedthepresenceofafranchiseasthedependentvariableandconducted
Page 11
10
multivariableregressions.Bruggink&ZamparelliandBaehrusedtheresultstoseewhich
geographiesweremostsimilartothosewithfranchisesalready.Theyusedthatresultto
determineratingsfornewfranchiselocations.Thisprovidedafoundationofproven
variablesusefulinpredictingthelocationoffranchisesandalsogaveinsightintothebest
waytostatisticallyanalyzethistypeofproblem.
Focusingonthepresenceofteamsinmarketsandusingthistogenerateresults
impliesanassumptionthattheexistingfranchiseswereallplacedindesirablemarketsand
thatthesamedecisionmechanismusedforplacingthesefranchisesshouldbeusedfor
placingfuturefranchises.Theapproachfailstolearnfrompreviousdecisions.Additionally,
byfocusingonjustthepresenceofateam,allteamsarevaluedequally.Forexample,in
soccer,asmentioned,theSeattleSoundersconsistentlyhavethehighestyearlyaverage
attendanceandthuscitiesmoresimilartoSeattlewouldbedesirableforexpansion.
However,usingtheabovedescribedapproach,thecharacteristicsofSeattlewouldbe
dilutedbythenoisecreatedbycitieslikeDenver,homeoftheColoradoRapids,oneofthe
leastsuccessfulfranchisesinMLS.Inshort,MLSwantsmoreteamslikeSeattle,yetby
focusingonwhatpredictsteampresence,citieslikeDenverwouldbeviewedasnomoreor
lessidealthancitieslikeSeattlebecauseanalysisfailedtoevaluateactualindicationsof
teamperformance.
iv.Summary
Consideringthesethemesandtheirshortcomings,itwasapparentthatthispaper
servedasabridgethatconnectedandintegratedthefindingsofprior,isolatedworks.This
paperholisticallyexaminedanumberoffactorsthathaverepeatedlybeenobservedtoplay
Page 12
11
asignificantroleindeterminingconsumptionandattendanceintheliterature,acrossa
varietyofstudiesandmarkets.Byconsideringthesefactorsjointly,andexploringtheir
effectsacrossanumberofgeographicmarkets,itbecamepossibletoreliablyidentifythe
characteristicsthatmakeanareaparticularlysuitableforMLSexpansion.Applyingthis
approachtoAmericanprofessionalsoccer,thepresentstudyexpandeduponthe“verylittle
research[which]hasbeendoneonthistopicofthelocationofexpansionteamsin
professionalsports”(Bruggink&Zamparelli,1999,p.50).
III.Methodology
i.FormulationofHypotheses
Toeffectivelyaddresstheresearchquestion,anumberofhypotheseswere
evaluatedviamultivariablelinearregressioninR.Eachofthehypothesesfocusedon
measuringamarket’spotential,ratherthantheabilityofmanagementtomaximizethat
potential,conditionalonentry.Dataregardingthefactorsneededtoevaluateeach
hypothesiscameentirelyfromtheUSCensusandUSConsumerExpenditureSurvey.
Accordingly,factorsunderthecontrolofteammanagement,suchasateam’son-field
success,wereincorporatedmerelyascontrolvariables;thesevariableswerenotthefocus
oftheanalysisbecausetheydidnotdeterminetheinherentviabilityofthemarket,and
couldnotbeusedtoinformdecisionsaboutwheretolocateanewfranchiseinthefuture.
DataregardingteamspecificvariablescamefromWikipediaandwasverifiedthroughthe
MLS’swebsite,teamwebsites,andotheronlineresources.
Eachhypothesisfocusesonfranchisesuccessasthedependentvariable;however,
thefactorsdiscussedinthehypotheseswerealsoexaminedinthecontextoftheirabilityto
Page 13
12
predictthepresenceofafranchise(presencemodel)inamarketandtheresulting
regressionmodelprovidedabenchmarktowhichtheregressionbasedonattendance
(attendancemodel)couldbecompared.
Hypothesis1:CommunityAttachment-Apositiverelationshipexistsbetweencommunity
attachmentandMLSfranchisesuccess.
Thefirstfactorconsideredwascommunityattachment.Iproposedthatmetroareas
withahighersenseofcommunityattachmentwouldbemoresuccessfuloverall.This
hypothesisdrawsfromresearchbyUhlman&Trail,whichfoundthatSeattleSoundersfans
citedcityprideandattachmentasamoresignificantfactorintheirloyaltytotheteamthan
theiractualpassionforthegameofsoccer,oreventhespecificteam.Thistheorywas
derivedfrompreviousworkdonebyAndersonandStone(1981),Funketal(2001),and
Mahony,Nakazawa,,Funk,James,&Gladden(2002).
Hypothesis2:TotalSports-Anegativerelationshipexistsbetweenthenumberofother
professionalsportsteamsinamarketandMLSfranchisesuccess.
Next,Iconsideredthepresenceofalternativeentertainmentoptions,which
intuitivelypossessedanegativeimpactonMLSfranchisesuccess.Kim&Trail(2010)found
thistobethecasewhentheyexaminedfactorsthataffectattendanceatawomen'
professionalbasketballgame.ThisheldweightparticularlyforMLSbecausebothMLSand
women'sprofessionalbasketballarerelativelyyoungprofessionalsports,whicharequite
likelytobeeclipsedbylargerleaguesinthesamemarket,specificallyteamsfromthebig
Page 14
13
fourleaguesofNorthAmericansports:theMLB,NBA,NFL,andNHL.Thepresenceof
teamsfromothersportsleagueswasexpectedtodetractfrominterestinanMLSfranchise
because,intheeyesofgeneralsportsfans,MLSisasubstitute,andpotentiallyevenan
inferiorgood.
AdditionalsupportforthishypothesiswasprovidedintheworkofFunketal.,
whichnotedthatsomeoftheappealofattendingmajorsportingeventswasattributedto
thenoveltyofwatchingsportsinperson.Themoreopportunitiestowatchprofessional
sportsinaparticulargeographicmarket,thelessattentionanMLSfranchisereceives.
Thereexistedsomeliteraturewhicharguedthathavingadditionalprofessionalsports
franchisesestablishedacityasa“sportscity.”Proponentsofthislineofthinkingargued
thatsupportingteamsbecomespartofthelocalidentity,leadingittobea“majorleague”
city(Bruggink&Zamparelli1999).Thismayholdmeritforlargerleagues;however,fora
youngleague,likeMLS,itseemedmorelikelythatafranchisewouldbeovershadowedby
otherprofessionalfranchises.
Hypothesis3:NeedforAesthetics-Apositiverelationshipexistsbetweentheneedfor
aestheticsandMLSfranchisesuccess.
AstudyconductedbyFunketal.(2012)indicatedthattherewasapositive
correlationbetweenconsumersneedforaestheticvalueandtheirconsumptionofsports.
Aestheticvalue,accordingtoFunketal."representsperformanceastheinherentdesirefor
aestheticrepresentationandmovement,desireforexcellence,beauty,andcreativityof
athleticperformance,fantasyandsubjectiveexperiencesthatarepleasurable"(359).In
Page 15
14
theirresearchitwasdeterminedthatthehighertheneedforaestheticsanindividualhad,
themorelikelytheyweretoconsumesports.
Hypothesis4:Population-ApositiverelationshipexistsbetweenpopulationsizeandMLS
franchisesuccess.
Anotherimportantfactortoconsideredwasthegeneralpopulationofametroarea.
Byhavingalargerpopulationinthemetroarea,therearemorepotentialattendees,andan
overallhighernumberofpotentialcustomers.Inpreviousprofessionalsportsexpansion
modelssuchastheoneputforwardbyBruggink&Zamparelli,populationsizewasatop
predictorofthepresenceofaprofessionalsportsfranchise.Populationwasalsoatop
predictorofMLSfranchiselocationsaccordingtoBaehr(2006).Althoughtheseprevious
modelsonlysoughttopredictthepresenceofateam,notateam’ssuccess,itwasstill
importanttoevaluatesheerpopulationinthiswork,becauseitwasquitelikelytopredict
MLSfranchiseattendance,andthussuccess.Additionally,populationwasimportantto
includeinapredictivemodelbecause,atsomepoint,itdoeslikelybecomealimitingfactor.
Thatis,insomemarkets,thepopulationissimplytoosmallforanMLSfranchisetobe
economicallyviable.Infact,populationaloneservesasareasonablepredictoroffranchise
presenceasillustratedinFigure3.
Page 16
15
Figure3:PresenceofanMLSfranchisebypopulationforallmetroareasintheUS
PanelA:PopulationDistributionofMSAswithanMLSFranchise
Page 17
16
PanelB:PopulationDistributionofMSAswithoutMLSfranchises
Asthecitiesgetlarger,youbecomemorelikelytofindanMLSfranchiselocatedinthatcity.Note:TheY-axischangesfromPanelAtoPanelBduetothehigheroverallvolumeofMSAswithoutfranchises.Hypothesis5:Knowledge-Apositiverelationshipexistsbetweendesiretoacquire
knowledgeandMLSfranchiseattendance.
Thefinalfactorevaluatedwasthelocalpopulation’sdesiretoacquireknowledge.
Thisvariablehadnotbeenincludedinotherprofessionalsportsfranchiseprediction
models;rather,itwasdrawnfromfanmotivationtheory.JamesandTrailcitedthe
Page 18
17
acquisitionofknowledgeasafactorthatplayedanintegralroleinfanbehavior.Fansoften
viewedsportsasahobbytobestudied,indicatingahigherdesiretoacquireknowledge
abouttheirteam.ThisoftenledtosupportofaspecificteamintheformofTVviewership
and,importantlyforthisstudy,gameattendance.
ii.DataCollection
Duetothescopeandnatureoftheresearchquestion,primarydatacollectionwas
waived.Fortunately,anumberofsecondarydatasourceswerereadilyavailablethat
addressedtheresearchquestion.Anumberofproxieswereidentified,operationalizingthe
constructsdiscussedinthehypothesisdevelopmentsection.First,averagesinglegame
attendancewasidentifiedasthemostsuitableproxyforfranchisesuccessandcollected
fromworldfootball.net3.SmallerleaguessuchastheMLSandWNBAtendtorelymore
heavilyonattendanceasasourceofrevenuewhilelargerleaguesrelyonTVcontracts
(KimandTrail,2010).Asmentionedintheliteraturereview,theuseofsomemeasureof
successasadependentvariabletoevaluatemultiplehypothesessetsthisresearchapart
frompreviousstudies.Themajorityofprevioussportsleagueexpansionmodelsassume
thatexistingfranchiselocationswereoptimalandthusthatfutureexpansionsitesshould
beassimilaraspossibletothoseexistingsites.Byfocusingonattendance,andcomparingit
tothepresenceofateam,thisstudyidentifiedpopulation-relatedfactorsthatassociated
withthesuccessofMLSfranchises,followingtheircreation.
32014tablebuteachyearfrom2005-2014wascollected.http://www.worldfootball.net/attendance/usa-major-league-soccer-2014/
Page 19
18
Proxiesforcommunityattachment,needforaesthetics,anddesireforknowledge
werealsoneeded.Comprehensivecommunityattachmentmeasuresformajor
metropolitanareasacrossAmericawerenotreadilyavailableandwouldnecessitate
extensivedatacollection,whichfelloutsidethescopeofthisproject.However,theUS
Census,themainsourceofdataregardingthefactorsexaminedinthisstudy,incorporates
dataonthenumberofindividualsthatliveinthesamestatewheretheywereborn4.This
measurewasusedasaproxyforcommunityattachment.Itstoodtoreasonthatindividuals
whowereborninaspecificstate,andultimatelydecidedtocontinuelivinginthatsame
geography,likelydevelopedadegreeofattachmenttotheircommunity.
Similartocommunityattachment,collectingprimarydataonthedesirefor
knowledge(asJamesandTraildidfortheirinitialstudy)inallmetroareasacrosstheUS
wouldhaveentailedagreatdealoftimeandeffort,andwasthusdeemedoutofscopefor
thispaper.However,asuitableproxyexistedinformofeducationalattainment5.Thus,
ratherthanemployaspecificmetricofdesiretoacquireknowledge,theanalysisdrew
uponameasureofthepercentageofthepopulationwithatleastabachelor’sdegree.This
datawasavailableonaMetropolitanStatisticalArea(MSA)levelfromtheUSCensus.Using
thisasaproxyforthedesireforknowledgefollowedthelogicthatifanindividualwas
motivatedtocompleteabachelor’sdegree,thentheyhavesomedesiretoacquire
knowledge.Certainlyfactorsotherthanthedesireforknowledgemotivate(andlimit)
42014tableasanexampleforrepeatability.http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid=ACS_14_5YR_B06001&prodType=table52014tableasanexampleforrepeatability.http://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?pid=ACS_14_1YR_S1501&prodType=table
Page 20
19
individualstopursuehighereducation,butoverallahigherpercentageofindividualswith
bachelor’sdegreesindicatesahigherpercentageofindividualsseekingoutknowledge.
Entertainmentspendingalsorequiredtheuseofaproxyandthusutilized
householdentertainmentspending.Thiswasasuitable,andevenpotentiallyadvantageous
proxyforanumberofreasons.First,itnotonlyrepresentedaconsumer’sdesiretoenjoy
aesthetics,butalsotheirabilityandwillingnesstopayforthosedesires.Thiscontrasted
previousstudieswhichexaminedtheimpactofhouseholdincomeonthepresenceof
sportsfranchises,anapproachthatfailedtoaccountforhowincomeisspent.Household
entertainmentspendingwascollectedfromtheUSConsumerExpenditureSurvey6.
MLSpointtotalsforeachseasonwerecollectedfromMLS.com7.Dataaboutthe
presenceofother‘bigfour’NorthAmericanprofessionalsportsfranchisesineachmarket
wasdrawnfromWikipedia8.
Withrespecttosampleselection,therewereeffectivelytwotypesofmarketsthat
wereconsidered.ThefirstwasthesampleofmarketsthatcurrentlyhostMLSteams.All
datawascollectedonthesemarkets(MSAs/franchises)fortheperiodbetween2005and
2014;asfarbackascensusdatawasreadilyavailable.Thisperiodsawafairlyconsistent
setofteams,onlyoneteamfoldedandfivewereaddedintheUS.Montreal,Toronto,and
Vancouveralsolaunchedoverthisperiodbutareexcludedfromthedataset.Dataonthose
franchiseslocatedinCanada(MontrealImpact,VancouverWhitecaps,andTorontoFC)
wereexcludedfromthestudybecauseofdatainconsistencyissues(e.g.,theCanadian
6http://www.bls.gov/cex/csxmsa.htm72014tableasanexampleforrepeatability.http://www.mlssoccer.com/standings/mls/2014/8https://en.wikipedia.org/wiki/U.S._cities_with_teams_from_four_major_league_sports
Page 21
20
governmentrecordsdataindifferentformats,etc.).Thesecondsetofmarketsthatneeded
tobeconsideredwereMSAsthatdidnothaveMLSfranchises–i.e.,potentialexpansion
locations.MSAswereexcludedfromtheattendancemodelprojectionsiftheirpopulation
waslessthanonemillionpeople,becausethiswastheapproximatesizeofthesmallest
metroareathatcurrentlyhostsanMLSfranchise;theSaltLakeCitymetroareaand
trainingofthemodeloccurredonadatasetwhereallMSAspossessedrelativelylarge
populationssousingthemodelonadissimilardatasetprovedproblematic.
Beyondthekeyindependentvariablesofinterest,anumberofcontrolvariables
werealsoconsidered.Throughresearchandintuition,itwasdeterminedthatthreefactors
mustbeaccountedforinanyanalysisofgameattendance:teamon-fieldsuccess,franchise
novelty,andleaguegrowth.Thesefactorsonlyapplytotheattendancemodelandnotthe
presencemodelastheydependonthepresenceofateaminaspecificcityandwereheld
constantinthepresencemodel.Inotherwords,ifcontrolvariablesenteredthepresence
model,theywouldholdhighpredictivevaluebutnotactuallyprovidemeaningfulinsights
becauseonlymarketswithateamhadvaluesforthosefactors.
Previousresearch,suchastheworkbyKimetal.andthatbyParrish(2013),
supportsthenotionthatwinningimpactsattendancegreatly.Teamperformancewas
thereforemeasuredasafranchise’spointspergame.Averagepointspergamewasused,
ratherthansheernumberofwins,orthewin-lossratio,ortotalpointsaccumulatedovera
season.ThiswasbecausemanyMLSgamesendinatie.Finalrankingsaredeterminedby
totalpointseachyear,withawinresultinginthreepoints,atieresultinginonepoint,anda
lossresultinginnopoints.Becausethenumberofgamesplayedbyeachfranchisevaried
overtime,withtheintroductionofnewteams,rawpointtotalsprovidedaninconsistent
Page 22
21
measureofperformance.Incontrast,averagepointspergamewasastable,consistent
measure,overtime.Theliteraturealsorevealedthatthenoveltyofafranchiseimpacted
supportforateam,soafranchiseagevariablewasalsoincluded(Baade,1994).Finally,
dummyvariablesindicatingyearswerealsoincorporatedintheanalysis,tocapturelarger
temporaltrends,suchaschangesintheleague’spopularityoverall.Table1containsa
summaryofeachhypothesisandthedatausedtoevaluateit.
Table1:HypothesesandMeasures
Notes:Thistablegiveseachhypothesisalabel,statesthehypothesisintermsoftheindependentvariables(nottheoriginalfactorsfromtheliterature),statestheexpectedrelationship,andprovidesabriefdescriptionofthesupportingargument.iii.Analysis
Oncethedatawasgathered,statisticalanalysiswasconductedintheformofthree
multiplelinearregressionmodels.Thefirstmodelwasananalysisofhowthecontrol
variablesimpactedattendance.Thecontrolvariablesinthisinstanceincludedpointsper
Page 23
22
game(on-fieldsuccess),dummyvariablescapturingeachyear,tocontrolfortimeeffects,
aswellasafranchiseagevariable.
Theformulaforthisinitialstatisticalmodelwasasfollows:
𝐴𝑡𝑡𝑒𝑛𝑑𝑎𝑛𝑐𝑒 = 𝛽* +𝛽-𝑃𝑜𝑖𝑛𝑡𝑠𝑃𝑒𝑟𝐺𝑎𝑚𝑒 + 𝛽5𝐹𝑟𝑎𝑛𝑐ℎ𝑖𝑠𝑒𝐴𝑔𝑒 + 𝛽92014𝐷𝑢𝑚𝑚𝑦
+ 𝛽92013𝐷𝑢𝑚𝑚𝑦…+𝛽-52006𝐷𝑢𝑚𝑚𝑦
Aftergainingabetterunderstandingontheimpactthatwinning,novelty,andyearly
changes9had,thefullmodelexpandedonthebaselinemodel,toincludeallindependent
variables.
𝐴𝑡𝑡𝑒𝑛𝑑𝑎𝑛𝑐𝑒 = 𝛽* +𝛽-𝑃𝑜𝑖𝑛𝑡𝑠𝑃𝑒𝑟𝐺𝑎𝑚𝑒 + 𝛽5𝐹𝑟𝑎𝑛𝑐ℎ𝑖𝑠𝑒𝐴𝑔𝑒 + 𝛽92014𝐷𝑢𝑚𝑚𝑦
+ 𝛽D2013𝐷𝑢𝑚𝑚𝑦 +⋯+𝛽-52006𝐷𝑢𝑚𝑚𝑦 +𝛽-9𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛
+ 𝛽-D𝐸𝑛𝑡𝑒𝑟𝑡𝑎𝑖𝑛𝑚𝑒𝑛𝑡𝑆𝑝𝑒𝑛𝑑𝑖𝑛𝑔 − 𝛽-K𝑇𝑜𝑡𝑎𝑙𝑆𝑝𝑜𝑟𝑡𝑠𝐹𝑟𝑎𝑛𝑐ℎ𝑖𝑠𝑒𝑠
+ 𝛽-M𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑛𝑆𝑡𝑎𝑡𝑒𝑜𝑓𝐵𝑖𝑟𝑡ℎ +𝛽-Q𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛
Eachofthebetasbeingtestedwasevaluatedbasedonp-valuesandwasheldtoa
significanceleveloffivepercent.Thesignofeachindependentvariablewasalsoadjusted
toalignwiththeanticipatedrelationshipbasedonthestatedhypotheses(i.e.anexpected
negativerelationshipgeneratesanegativecoefficient).Theonlyvariablethisimpactedwas
β14whichpertainedtothepresenceofotherprofessionalsportsteamswhichwas
hypothesizedtohaveanegativerelationshipwithattendance.Oncethefullattendance
modelwasevaluated,itprovidedcoefficientsthatwereusedtodeterminewhichpotential
expansionsiteswouldbebestforMLSfranchises.Boththisandthecontrolmodelwere
trainedonthesamplecontainingexistingMLSfranchises.
9Theimpactoffranchiseagesquaredwasalsoconsideredbutultimatelyremovedastherelationshipbetweensuccessandageappearedtobelinear.
Page 24
23
Finally,IgeneratedathirdmodelbasedonthepresenceofanMLSfranchise.This
thirdmodelreliedonthecompletesetofMSAsforatrainingdatasetandtheresulting
modelpredictedthelikelihoodofeachMSAhavinganMLSfranchise(presenceofa
franchise).Thepresencemodelremovedthecontrolvariablesfromtheattendancemodel
becausetheydidnotimpactthelikelihoodofafranchiseexisting,butratherthesuccessof
thatfranchise.Also,allpredictionsforthepresencemodelutilized2014dataleadingtothe
removaloftheyeardummyvariables.Thisresultedinthefollowingmodel:
𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒 = 𝛽* + 𝛽-𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽-𝐸𝑛𝑡𝑒𝑟𝑡𝑎𝑖𝑛𝑚𝑒𝑛𝑡𝑆𝑝𝑒𝑛𝑑𝑖𝑛𝑔
− 𝛽5𝑇𝑜𝑡𝑎𝑙𝑆𝑝𝑜𝑟𝑡𝑠𝐹𝑟𝑎𝑛𝑐ℎ𝑖𝑠𝑒𝑠 + 𝛽9𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑛𝑆𝑡𝑎𝑡𝑒𝑜𝑓𝐵𝑖𝑟𝑡ℎ
+𝛽D𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛
Thispresencemodeloriginallygeneratedcuriousresults,withmanysmallercities
receivinghigherscoresthanMSAsthatintuitivelyseemedtobetterfittheexpectationsfor
anMLSexpansionlocation.Uponfurtherinspection,thepresencemodelgenerateda
negativecoefficientforpopulation(Table2)asaresultofthecovariancebetween
populationandtotalsports(Table3).
Table2:OriginalPresenceModelRegressionResults
Page 25
24
Note:Populationhadanegativecoefficientintheoriginalmodelresultinginmanylarger
citieslackingotherprofessionalsportsbeingconsideredpoorlocationsforMLSfranchises.
Table3:CorrelationofOriginalPresenceModel
Theattendancemodelalsoexhibitedthisbehaviorbutduetoitonlyconsidering
MSAswithpopulationsoveronemilliontheimpactwaslesspronounced.Asaresult,total
sportswasultimatelyremovedfromthepresencemodelleavingthemodelbelowasthe
completepresencemodel.
𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒 = 𝛽* + 𝛽-𝑃𝑜𝑝𝑢𝑙𝑎𝑡𝑖𝑜𝑛 + 𝛽-𝐸𝑛𝑡𝑒𝑟𝑡𝑎𝑖𝑛𝑚𝑒𝑛𝑡𝑆𝑝𝑒𝑛𝑑𝑖𝑛𝑔
+ 𝛽9𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑛𝑆𝑡𝑎𝑡𝑒𝑜𝑓𝐵𝑖𝑟𝑡ℎ +𝛽D𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛
iv.Assumptions
Theresultsoftheattendancemodelandthecontrolmodelwerecomparedto
evaluatewhichmetroareasprovidethemostopportunityforMLSexpansion.Inorderto
dothis,thetheoreticalnewfranchisesinthemodelswereassumedtohave1.39pointsper
game(theMLSaveragein2014),andwereassumedtobeanewfranchisethatyear,
meaningtheirfranchiseagewas0.
IV.Results
Page 26
25
Oncealldatacollectionwascompleted,itwasanalyzedusingtheRprograming
language.Specifically,thethreemultiplelinearregressionsdescribedinthemethodology
sectionwerecreatedtoaddresswhatfactorswereimportantindeterminingtheviabilityof
anMLSmarketasmeasuredbyattendanceandbypresence.Thecoefficientsfromthefull
(controlvariablesandindependentvariables)attendanceregressionmodelwerethenused
toproject2014attendanceforpotentialMLSexpansionlocations,andtheindependent
variablesfromthepresencemodelwereusedlikewise.Thissectiondetailstheresultsof
thoseanalyses.
i.EvaluationofHypotheses
Table4:SummaryofHypothesisTests
Table4showstheresultsofamultiplelinearregressionranonallcontrolvariablesandallindependentvariables.“***”indicatesasignificance<.001,“**”<.01,“*”<.05,and““>.05. Thefivehypothesesposedbythisstudywereevaluatedbycreatingamultivariable
linearregressionmodelwithattendanceasthedependentvariable.Findingsindicatedthat
apositiverelationshipexistedbetweenpopulationsizeandMLSfranchisesuccess.Results
alsoindicatedastatisticallysignificantrelationshipbetweenattendanceandpopulation.
Foreachadditionalonemillionpeopleinametroarea,ateamsawapproximatelyan850
personincreaseinpergameattendance.However,forlargercities,countertothisimpact,a
Page 27
26
negativerelationshipexistedbetweenthenumberofotherprofessionalsportsteamsina
marketandMLSfranchisesuccess.Essentially,althoughlargerpopulationstendtoleadto
moresupport,theyalsotendtoattractmorecompetitionintheformofothersports
franchises,eachofwhichreducedprojectedattendancebyapproximately1,300pergame.
Theresultsfurtherindicatedthatapositiverelationshipexistedbetweenspending
onentertainmentandMLSfranchisesuccess.Metroareasthatspentmoreon
entertainmentsawattendanceriseapproximately4peoplepereveryadditionaldollar
spentonentertainmentonaverage.Supportalsoindicatedapositiverelationshipbetween
educatedshareofthepopulationandMLSfranchisesuccess.MSAswhereagreater
percentageofindividualscompletedaleastabachelor’sdegreestendedtobettersupport
MLSfranchises.
Countertoexpectations,testingrevealedanegativerelationshipexistsbetween
communityattachmentandMLSfranchisesuccess.Essentiallytheresultsindicatedthat
thehigherpercentageofpeoplelivingintheirplaceofbirth,thelesstheysupportedan
MLSfranchise.
Asapointofcomparison,eachhypothesisincludedinthepresencemodel
(population,needforaesthetics,desiretoacquireknowledge,andcommunityattachment)
resultedinacoefficientwiththesamesign,signifyingacomparablerelationshipwith
presence.However,forthepresencemodelonlypopulationanddesireforknowledge(as
educationalattainment)resultedinstatisticallysignificantfindings(AppendixA).
Comparingtheresultsofthecontrolmodel,theattendancemodel,andthepresencemodel
yieldedTable5belowwhichcomparestheresultingbetavalues(coefficients)and
significancesofeachfactorconsideredineachmodel.Table8alsoprovidestheR2value
Page 28
27
thatresultedforeachmodelwiththeattendancemodelholdingthehighestvaluefollowed
bythepresencemodelandlastlythecontrolmodel.
Table5:Comparisontableofresultsofthecontrol,attendance,andpresencemodels
Notes:.“***”indicatesasignificance<.001,“**”<.01,“*”<.05,and““>.05.
ii.Controlmodelresults
Thetablebelow(Table5)outlinesthegenericresultsofthecontrolmodelwhich
includeddummyvariablesforeachyearinthedataset,theageofthefranchiseinthatyear,
andthepointspergameassociatedwiththefranchiseforthatyear.Forthefullresults,see
AppendixB.
ThecontrolmodelhadanR2valueof.3711,indicatingthatitexplained
approximately37%ofthevarianceinattendance.Thisprovidedabaselinetowhichthefull
attendancemodeliscompared.
Page 29
28
Table6:SummaryofControlModelResults
Notes:Table2showsthatpointspergame(on-fieldsuccess),franchiseage(novelty),andsomeyears,didhaveasignificantimpactonattendance.“***”indicatesasignificance<.001,“**”<.01,“*”<.05,and““>.05.
Thecontrolmodelalsoshowedthatseveralofthecontrolvariablesweresignificant,
includingpointspergameandfranchiseage.ItalsoshowedthatMLShasseenyearover
yeargrowthwitheachyearhavingapositiverelationshipwithattendancetovarying
degreesofsignificance.Theseresultssupportedpreviousresearchwhichindicatedthat
noveltyandfranchiseon-fieldsuccesshaveanimpactonsportsconsumptionandfan
motivation.Thismodelalsogeneratedpredictedattendancevaluesfor2014forexisting
MLSlocations(Table6)wherewerecomparedtoactualattendancenumbersasacheckof
theaccuracyofthemodel.
Page 30
29
Table7:ControlModelPredictedAttendancevsActualAttendancefor2014
iii.Fullmodelresults
Thefullregressionmodelbuiltoffthecontrol.Thefullattendancemodeloverallhad
anR2valueof.527,indicatingthatitexplainedanadditional15%ofvariationinattendance
comparedtothecontrolmodel.Forthefullresults,includingexactcoefficientsand
significancevalues,aswellastheresultsofthecontrolvariablesinthefullmodelsee
AppendixC.Asacheckofaccuracy,theresultingmodelgeneratedpredictedattendance
valuesforthe2014seasonwhichwerecomparedtotheactualattendancefor2014as
showninTable7.
Page 31
30
Table8:FullModelPredictedAttendanceVsActualAttendancefor2014
iv.Projected2014attendanceandpresenceforpotentialexpansionlocations
Oncecoefficientvariablesweredeterminedthroughthefulllinearregressionmodel,
thosecoefficientswereusedtoestimateattendancevaluesforpotentialexpansion
locationsforMLSfranchises.Aspreviouslymentioned,potentialMLSexpansionlocations
werelimitedtoUSmetropolitanstatisticalareaswithpopulationsoveronemillionpeople.
TheresultsareshowninAppendixDandtheprojectedattendancevaluesinthetable
representtheexpectedattendanceforanexpansionfranchise,soformetroareasthat
alreadyhaveMLSfranchises,thisrepresentsasecondfranchise.Asapointofreference,the
averageMLSattendancefor2014was19,148.Utilizingthecoefficientsgeneratedbythe
presencemodel,predictedpresencevalueswerealsogeneratedforallMSAsintheUS,
theseresultsareshowninAppendixE.Tocomparetheresultsofthemodels,Figure4
showsascatterplotoftheprojectedpresencemodelresultsagainsttheprojected
attendancemodelresults.Theseresultsarelimitedtopopulationsoveronemillionbutthe
fullresultscanbeseeninAppendixF.
Page 32
31
Figure4:ProjectedPresencevsProjectedAttendanceforMSAswithPopulations
overonemillion
Figure4illustratesthedifferencebetweenthetwomodels.Notably,thetopten
resultsfromtheattendancemodelplaceabovelocationswithfranchises,andalltenof
theseMSAsscoredbelowa.5inthepresencemodel,butstillrankamongthetoplocations
basedonthatmodelaswell.
Page 33
32
V.Discussion
Thissectionwilldiscussinterestinginsights,newquestionsforresearch,and
opportunitiesforimprovinguponthisstudy.
i.AttendanceModel
ConsideringthecomparisontotheactualattendancedatashowninTable7,the
attendancemodelprovidesafairlyreliablepredictivetool.Seattlecreatedthelargest
discrepancybetweenpredictedandactualasexpectedduetoitsstatusasanoutlier
comparedtotypicalMLSattendance.Thatsaid,eachofthemodelscreatedpossessed
strengthsandweaknesses.Asintended,theattendancemodelprovidedaninsightintothe
potentialsuccessofanewfranchise,notjusttheabilitytosaywhereateamshouldor
shouldnotexist.Thecorrelationbetweenpopulationandtotalsportsmaydistortthe
modelsome(Table8),butthisdidnothaveaslargeofanimpactontheattendancemodel
asonthepresencemodelandIfeltthatdespitethecovariance,bothwereimportant
factorstoevaluate.
Table9:Fullmodelindependentvariablecorrelationtable
Note:Totalsportsandpopulationwerehighlycorrelatedbutultimatelywerebothincludedinthefullattendancemodel.
Page 34
33
Also,duetobeingtrainedonadatasetmadeupofonlylargecities(thosewithMLS
franchisesalready),theattendancemodelgeneratedrelativelyasmallcoefficientfor
population.Thistendedtobiasthemodeltobemoreforgivingtowardssmallercitiesin
comparisontothepresencemodelandbecauseofthis,hadInotfilteredoutpopulations
smallerthanonemillion,Boulder,ColoradowouldhavebeenthetoplocationforMLS
expansionprojectingnearly37,000inaverageattendancedespitetheMSAonlyhavinga
populationofapproximately350,000.
Further,themajorityofprojectedexpansionlocationshavehigherprojected
attendancesthancurrentMLSteam.Thisresultedpartiallyduetoeachofthesefranchises
beingnewandnotfeelingthenegativeimpactoffranchiseage,butalsoindicatesthatthe
modellikelytendedtooverestimateattendances.Thenegativeimpactofagemayhave
resultedfromtheimpactofnoveltybutitalsomayhaveresultedearlyMLSfranchise
locationsnotbeingwiselyselectedormanaged.In2014fourofthefiveleastattended
franchiseswerefoundingmembersoftheleague.Thefifthofthosefranchises,Chicago,is
alsoamongtheolderteamsintheleague,joiningtwoyearsaftertheleague’sfounding.This
hasseriousimplicationsforMLSandforstrugglingfranchisesasitprovidedevidence
suggestingsomefranchisesshouldeitherberelocatedorfoldedinfavorofbetterlocations,
oratleastnewerfranchisesshouldMLScontinuetostrugglefinancially.Thisraisesthe
questionofdidfranchiseageanddecreasingnoveltyreallydrivedownMLSattendance,as
suggestedbythemodel,orweretheoriginalfranchisesjustpoorlychosen?
ii.PresenceModel
Page 35
34
Thepresencemodelinthisstudyservedasabenchmarktowhichtheattendance
modelcouldbecompared.Countertotheattendancemodel’sbiastowardssmallcities,the
presencemodelreflectsbiastowardsgenerallylargercitiesbecauseitwasestimatedona
datasetcontainingallMSAswherepopulationvariedmoredramatically.Thismodel
appearedtomorecloselyreflecttheapproachMLShastaken,witheightofthetopten
predictedmarketsalreadyhostingfranchisesandwithMiamiandAtlantaabouttomakeit
tenoutoften.
iii.Controlvariablesplayabigpartinattendance
Oneinsightgainedfromtheresultsshowedthatthecontrolvariablesexamined
explainedafairamountofthevariationinattendanceinMLS.Thisisinterestingbecauseit
meansthatwithallreasonablemarkets(heredefinedasmetroareaswithpopulationsover
onemillion)thereisacertainbaselineattendanceyouwouldexpecttosee.Italsomeans
thatafairamountofthefranchise’ssuccesscomesdowntoeffectivemanagementwhich
wasmostclearlyrepresentedbythelargeimpactthatwinninghadonMLSattendance.
ReferringbacktoTable8,KansasCityembodiedthisfact.ForyearsSportingKansasCity
(previouslytheKansasCityWizards)struggledtoattractfansbutinrecentyearsthe
team’ssuccess(includingachampionshipin2013)pullsinfansleadingtooneofthetop
attendancemarksinMLS.ThePhiladelphiaUnionshowedtheoppositeimpactastheir
marketseemedpoisedforsuccessbutyearsofapooronfieldperformancehaslikely
drivendownattendance.Understandingtheimpactofmultipleyearsofon-fieldsuccessor
failureprovidesanavenueforfuturestudiestoexplore.Althoughotherfactorsplayedan
Page 36
35
importantroleinpredictingattendance,thesefindingsalsolendsomecredittothefamous
quotefromAlDavis,formerOaklandRaidersownerandGeneralManager;“Justwin,baby!”
iv.ModelsagreeMLSisnowmakingsolidexpansiondecisions
Despitehavingnoformallyexpressed,researchbasedmethodforevaluatingMLS
expansionlocations,theleagueseemstohaveagoodunderstandingofwhatmakesa
marketattractive.MLSfranchiseshavebeenannouncedforAtlanta,Minneapolis-St.Paul,
andMiamiwhichrepresentedthreeofthetopfiveprojectedcitiesforMLSattendance,and
threeofthetopeightcitiesforpredictedpresenceinMSAswithoutfranchises.
Additionally,theseresultsalsoprovideAtlantaFC,MinnesotaUnited,andthefuture
Miamifranchisearoughestimateofhowmanyfanstoexpectintheirfirstyearandcould
beusedtodetermineidealcapacitiesforstadiums.InMinnesotaUnited’scase,they
recentlyannouncedplansfora20,000seatstadiumbutthemodelsuggeststheycouldfilla
stadiumsignificantlylarger,althoughprojectedattendancefromthismodelappears
somewhatinflated.
v.Modelcouldbeexpandedtoincludemorevariablesforhigheraccuracy
Withboththecontrolandindependentvariables,therecertainlyexistedmore
variablesthatcouldbemeasuredandincludedinthemodelandthiswouldhavedrivenup
theR2value(perhapsjustartificiallybasedonthemechanicsofhowlinearregression
works)toexplainmoreofthevarianceinattendance.Thislistofvariableswasdeliberately
ashorterlistwithanemphasisonvariableswithsupportinexistingliteraturetoavoid
artificiallyinflatingtheR2value,butthelistofvariablesonecouldexploreinbuilding
Page 37
36
similarmodelsisalmostunlimited.Anotherreasonthislistofvariableswasusedisthat
therewaslittlecorrelationbetweenvariablesotherthanpopulationandnumberof
professionalsportsteamsinthatmetroarea.Otherfactorsthatwereinitiallyincludedbut
eventuallyremovedfromconsiderationduetolackofsupportinliteratureorhigh
correlationincludedaverageage,ethnicmakeup,stadiumage,andhouseholdincometo
nameafew.
vi.Residencyinstateofbirthapoorproxyforcommunityattachment
Basedontheresults,oneareathatshouldberevisitedinfutureworkishowto
evaluatecommunityattachmentbecauseresidencyinstateofbirthseemedtodoapoorjob
ofmeasuringthatsentiment.Communityattachmentwasafactorthatcamethroughthe
literatureasbeingoneofthemoststronglyinfluentialfactorsindeterminingMLSfan
motivation,sotoseeitnegativelycorrelatedwithMLSattendanceintheresultsseemed
unlikely.Itispossiblethattheunderlyingassumptionwasincorrectandinfactindividuals
whochoosetoliveinastatedifferentfromwheretheywerebornhaveagreatersenseof
attachmentbecausetheychosethatcommunitydespitehavingrootsinanotherlocation.
Whateverthecausemaybe,theresultseemedtobenotthatcommunityattachment
negativelyimpactsMLSattendance,butratherthatcommunityattachmentwasnot
accuratelymeasured.
VI.Conclusion
Page 38
37
Thissectionservestoreinforcethemaintakeawaysoftheresearch,illustratethe
limitations,synthesizethenewknowledgewhilecontinuingopeningupnewareasfor
exploration.
i.Summaryofkeytakeaways
Theresearchandresultsusedinthispaperhaveseveralstrengthsworthnoting
withimportantimplicationsgoingforward.First,thisstudyusedpubliclyavailabledata
andexistingresearchresearchbutapplieditinanewwaytohelpexpandtheliterature
aroundsportsexpansionmodelingand,inabroadersense,thewaybusinessesthingabout
enteringanewmarket.
Additionally,thestudyprovidedaframeworkthatitcouldbeappliedtoother
professionalsportsleagues,andotherbusinesses,lookingforwaystodetermineideal
expansionlocations.Itwouldbeparticularlyusefultosmaller,youngerprofessional
leagues,orfranchises,thatmaybeviewedasinferiorintheeyesofthegeneralpublicwhen
comparedtomoreestablished,traditionalbrandsliketheNFL,MLB,NHL,andNBAin
soccer’scase.Additionally,thevalidityandaccuracyofthisstudywouldbeeasytotestas
allinformationusedwaspublicallyavailableandrequiresnoprimarydatacollection.That
said,thequalityofthestudywouldbegreatlyenhancedwithwidescaleprimarydata
collection.Additionally,thehypothesesproposedarerelativelygeneralizableandrequire
littleifanyexpertiseinthefieldofsportswhichmeansmoreindividualswillbeableto
utilizethisresearchintheirownwork,evenifitisinanotherfield.
Also,theuseofmultiplelinearregressionallowedfortheanalysisofmultiple
variablesatonce,whileprovidinginformationabouthowthosevariablesworktogetheras
Page 39
38
well.Examiningsimplyonevariableinasimplelinearregression,orevenutilizingmultiple
simplelinearregressionswouldbeanoverlynarrowapproachandwouldbelessefficient
atminimizingthenoiseinthedata.Byputtingallofthevariablesintoonemodel,it
informedthemodelabouthowthosevariablesworkedtogethertobuildamorecomplete
pictureofhowtomodelattendanceforMLS.
Further,itisimportanttokeepinmindthatalthoughthisresearchprovideda
frameworkforfutureresearch,eachexpansionquestionwillhaveitsownnuancesandits
ownimportantvariables.WhatistrueforMLSmaynotbetrueforselectingMcDonalds
expansionlocations,andmaynotevenbetrueforselectingothersportsexpansion
locations.Duetothis,itremainsimportanttoalwaysstartwiththeliteraturetounderstand
whatmaybemostsignificantindeterminingthesuccessofexpansionlocationsinaspecific
industry,thenusetheavailabledatatodeterminethedegreeandsignificanceoftheeffect.
ii.Studylimitations
Themethodologythatwasusedinthisstudyalsohadmanylimitations.First,
proxieshadtobeusedtoevaluatesomeofthevariablesbeingstudiedduetolimitationsin
thedataavailable.Specifically,theproxyusedforcommunityattachmenthaslittlegrounds
intheliterature,butcommunityattachmentwasanewideaworthtestingandthiswasthe
mostviableoptionwiththedataavailable.Therewerealsosomelimitationstothe
entertainmentspendingdata.TheConsumerExpendituresurveydidnotcollectdataonall
metroareasbutratheronaselectfewfromeachoftheirfourpredefinedregions.These
metroareasprovidedvariedslightlyyeartoyearbutforexample,the2014metroareas
wereChicago,Detroit,Minneapolis-St.Paul,Cleveland,NewYork,Philadelphia,Boston,
Page 40
39
WashingtonD.C.,Baltimore,Atlanta,Miami,Dallas–FortWorth,Houston,LosAngeles,San
Francisco,SanDiego,Seattle,andPhoenix.Forcitiesnotlistedspecifically,theirregional
averagewasused.Asisevidentbythelistprovided,mostMLSmetroareaswereprovided,
butthisdatasetwasalargerlimitationonprojectingexpansioncitiesandlikelygenerated
regionalbiasesbasedontheneedtouseregionalaverages.
Additionally,thestudyonlyaccountedforMLSfranchisesbasedinAmerica.Thisisa
limitationbothontheoriginalmodel’saccuracyandontheabilitytoselectthebest
possiblecitiesforexpansion.Theoriginalmodelwaslimitedbynotbeingabletoinclude
severalyearsofdataforTorontoFC,VancouverWhitecaps,andMontrealImpact.
Additionally,metroareassuchasEdmonton,whichhasstrongsupportforalowerdivision
soccerfranchise(EdmontonFC),wereleftoutduetothedatainconsistenciesCanadian
teamswouldintroduce.
Further,thelackoftransparencyinMLS,andprofessionalsportsoverall,with
regardstofinancialreporting,providedalimitingfactor.Attendancewasusedasa
dependentvariabletomeasurethesuccessofafranchisehoweverattendancedoesnot
necessarilymeanprofitabilityforateam.FactorssuchasTVdeals,merchandise,ticket
prices,andothersallcontributetotheoverallprofitabilityandsustainabilityofafranchise.
Formostlargerprofessionalsportsleagues,Forbesalsoprovidesyearlyvaluationsalong
withrevenueandprofitabilityreports,howeverthesewereonlyavailableforthreeyears
forMLSatthetimeofthisstudy,whichprovidedlimitationsofitsown.Instead,raw
attendancedatawasusedandforsmallersportsleaguesliketheMLS,thisrepresentsa
morereasonableapproachtotakebecausethemajorityoftheirrevenuedoescomefrom
attendance(Kim&Trail,2010).Thatsaid,MLSnowoperatesunderanewTVdealthrough
Page 41
40
theircollectivebargainingagreementthatwilldramaticallyincreasethepercentageof
revenuecomingfromTV,thusthisapproachmayneedtobealteredinfuturestudies,but
didnotimpactthisstudyasittookeffectbeginninginthe2015season.
Finally,thisstudyattemptedtoaddressthelimitationsofpreviousworkthathad
studiedthepresenceofafranchisebyusingsomemeasureofsuccess.Bymakingthis
decision,itintroducedanewlimitationintothestudy.Becauseattendancebecamethe
dependentvariable,thesamplestudiedfortheattendancemodelconsistedofonly
locationsthathadMLSfranchises.Itispossibletheselocationshadsomethingaboutthem
thatmadethemsystematicallybetterlocationsandthatfactormaynothavebeencaptured
bytheattendancemodel.ThisatleastpartiallymanifestsitselfintheneedtofilterMSAs
withlowpopulationsbecausethemodelwastrainedondatacontainingonlyhigh
populations(thuslimitingthevarianceofthatfactor).Inandefforttoaddressthis,athird
modelwascreatedthatutilizedthesamevariablesasthepresencemodel(egnocontrol
variablesandremovingtotalsports),butusedattendanceasthedependentvariable,and
wastrainedonthecompletesetofMSAs.DuetothehighnumberofMSAshavingno
attendanceatall,thepredictedattendancevaluesasawholewerequitelowhoweverthis
providedanadditionalrankingsystemforevaluatingthequalityofexpansionlocations.
InterestinglyenoughthismodelalsogeneratedMiami,Atlanta,SanFrancisco,andPhoenix
asthemostappealingexpansionlocations,withMinneapolisratinghighlyaswell(Table9).
Thisonlyfurtherstrengthenedtheclaimothermodelsmadethatalsoprojecttheseastop
locations.
Page 42
41
Table10:AttendanceModelTrainedonAllMSAs
Note:MSAsthatdonotcurrentlyhaveMLSfranchisesareboldedaspotentialexpansionlocations.iii.Closingthoughts
Althoughthestudydidhaveanumberoflimitations,thisstudyalsocontributedto
theexistingliteratureinameaningfulway.Someresearchhadbeendoneintoevaluating
Page 43
42
expansionlocationsforprofessionalsportsleagues,butincomparisontoothersports
managementresearchquestions,theresearchbasewasrelativelythin.Bytakinganother
lookatthisquestionwithnewdata,newresearchvariables,andaslightlyaltered
approach,thisresearchcontinuedtopushforwardtheunderstandingofwhatmakessome
citiessupportprofessionalsports,andothersignorethem.Thisresearchalsovalidated
recentdecisionsmadebyMajorLeagueSoccerwhileleadingtoquestionsaboutolder
expansionlocations.Italsoopenedthefloorexaminewhatothervariablesnotincludedin
thismodelmightimpactMLSattendance.Finally,thisresearchcanservesoccerfans
locatedinmarketswithoutMLSteamsassomethingtheycanpointtoastheytryto
persuadepotentialteamowners,andtheleagueitself,tobringthebeautifulgametotheir
city.
Page 44
43
AppendixA:PresenceModelResults
Page 45
44
AppendixB:ControlModelResults
Page 46
45
AppendixC:FullModelResults
Page 47
46
AppendixD:ProjectedAttendanceForTopCandidateMSAs
Note:MSAsthatdonotcurrentlyhaveMLSfranchisesareboldedaspotentialexpansionlocations.
Page 48
47
AppendixE:ProjectedPresenceValuesfortopMSAsin2014
Note:MSAsthatdonotcurrentlyhaveMLSfranchisesareboldedaspotentialexpansionlocations.SaltLakeCity,homeofRealSaltLake(acurrentMLSfranchise)isnotshowninthistablebecauseitrankedsolow.Itspredictedpresencevaluewasapproximately.07.
Page 49
48
AppendixF:ProjectedPresencevsProjectedAttendanceforallMSAs
Page 50
49
WorksCited
Anderson, D. F., & Stone, G. P. (1981). Sport: A search for community. Sociology of sport: Diverse perspectives, 164-172.
Baade, R. a. (1994). Stadiums, Professional Sports , Economic Development: Assessing the Reality. Policy Study, 62(4), 1–39.
Baehr, M. J. (2006). An Analysis of Expansion and Relocation Sites for Major League Soccer. International Journal of Sport Managment., 7(January), 1–12.
Bruggink, T. H., & Zamparelli, J. M. (1999). Sports Economics: Current Research. (J. Fizel, E. Gustafson, & L. Hadley, Eds.). Westport, CT: Praeger Publishers.
Gitter, S. R., & Rhoads, T. A. (2010). Determinants of Minor League Baseball Attendance. Journal of Sports Economics, 11(6), 614–628. http://doi.org/10.1177/1527002509359758
Kim, Y. K., & Trail, G. (2010). Constraints and motivators: A new model to explain sport consumer behavior. Journal of Sport Management, 24, 190–210. Retrieved from http://www.scopus.com/inward/record.url?eid=2-s2.0-77955114752&partnerID=40&md5=b6c376860b51883bbf5ced7ee28f569f
Baehr, M. J. (2006). An Analysis of Expansion and Relocation Sites for Major League Soccer. International Journal of Sport Managment., 7(January), 1–12.
Bruggink, T. H., & Zamparelli, J. M. (1999). Sports Economics: Current Research. (J. Fizel, E. Gustafson, & L. Hadley, Eds.). Westport, CT: Praeger Publishers.
Davis, M. C. (n.d.). Income and the Locations of AAA Minor League Baseball Teams. University of Missouri-Rolla.
Gitter, S. R., & Rhoads, T. A. (2010). Determinants of Minor League Baseball Attendance. Journal of Sports Economics, 11(6), 614–628. http://doi.org/10.1177/1527002509359758
Funk,D.C.,Beaton,A.,&Alexandris,K.(2012).Sportconsumermotivation:Autonomyandcontrolorientationsthatregulatefanbehaviours.SportManagementReview,15,355–367.http://doi.org/10.1016/j.smr.2011.11.001
James,J.D.,Kolbe,R.H.,&Trail,G.T.(2002).PsychologicalConnectiontoaNewSportTeam :BuildingorMaintainingtheConsumerBase ?SportMarketingQuarterly,11(4),215–226.
Kim,Y.K.,&Trail,G.(2010).Constraintsandmotivators:Anewmodeltoexplainsportconsumerbehavior.JournalofSportManagement,24,190–210.Retrievedfromhttp://www.scopus.com/inward/record.url?eid=2-s2.0-77955114752&partnerID=40&md5=b6c376860b51883bbf5ced7ee28f569f
Page 51
50
Mahony,D.F.,Nakazawa,M.,Funk,D.C.,James,J.D.,&Gladden,J.M.(2002).MotivationalfactorsinfluencingthebehaviourofJ.Leaguespectators.SportManagementReview,5(1),1-24.
Parrish,C.(2013).Soccerspecificstadiumsanddesignatedplayers:ExploringtheMajorLeagueSoccerattendanceassumption.InternationalJournalofSportManagementRecreationandTourism,12,57-70.
Uhlman,B.T.,&Trail,G.T.(2012).ConsumerBehaviorAnAnalysisoftheMotivatorsofSeattleSoundersFCSeasonTicketHolders :ACaseStudy.SportMarketingQuarterly,21(4),243–252.