Effects of soccer on stock markets: the return-volatility relationship

(1)

ContentslistsavailableatSciVerseScienceDirect

The

Social

Science

Journal

jo u rn al h om epa g e :w w w . e l s e v i e r . c o m / l o c a t e/ s o s c i j

Effects

of

soccer

on

stock

markets:

The

return–volatility

relationship

夽,夽夽

M. Hakan

Berument

a,∗

_,

_Nildag

_Basak

_Ceylan

b

a_Department_of_Economics,_Bilkent_University,_06800,_Ankara,_Turkey

b_Department_of_Banking_and_Finance,_Yildirim_Beyazit_University,_06040,_Ankara,_Turkey

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received25March2011

Receivedinrevisedform11March2012 Accepted12March2012

Availableonline16September2012

JELclassiﬁcation: A12 C22 G12 L83

a

b

s

t

r

a

c

t

Thispaperassessestheeffectsofdomesticsoccerteams’performancesagainstforeign rivalsonstockmarketreturnsaswellasonthereturn–volatilityrelationship.Datafrom Chile,Spain,TurkeyandtheUnitedKingdomsupportpropositionsthatsoccerteamsresults ininternationalcupsaffectstockmarketreturnsandthereturn–volatilityrelationship. EvidencefromSpainandtheUK,soccerpowerhouses,suggeststhatlossesareassociated withlowerreturnsandhigherriskaversionbutevidencefromChileandTurkey,where socceristhemostimportantsportbutteamsarenotassuccessful,revealsthatwinsare associatedwithhigherreturnsandlowerriskaversion.

1. Introduction

Afundamentalassumptionofeconomicsisthatagents

arerationalindecisionmaking.Recentliterature,however,

haschallengedthisassumption.Stracca(2004)reviewsa

largebody ofevidence andﬁnds examplesof irrational

behaviorinnumerousstudies.Behavioraleconomicsand

behavioralﬁnancetrytoexplainhowemotionsand

cogni-tiveerrorsinﬂuenceinvestorsandtheirdecision-making

processes. Researchers can explain various stock

mar-ketanomalies,bubblesandcrashesusingpsychologyand

otherrelevantsocialsciencesmethods.Asthepsychology

literature suggests, “mood” is one of the sources of

夽 AlthoughthesportiscalledsoccerinNorthAmerica,itiscalled foot-ballintherestoftheworld.

夽夽 WewouldliketothankthestaffoftheEmbassyoftheTurkish RepublicinSantiago,ConsulateofChileinAnkara,IbrahimKirkayakof theTurkishRadioandTelevisionsportsservice,anonymousreferee,Rana NelsonandOnurIncefortheircontributions.

∗ Correspondingauthor.Tel.:+903122902342;fax:+903122665140. E-mail addresses: berument@bilkent.edu.tr (M.H. Berument),

nbceylan@ybu.edu.tr(N.B.Ceylan).

URL:http://www.bilkent.edu.tr/berument(M.H.Berument).

agent irrationality.FollowingEdmans, Garcia, and Norli

(2007)and Berument,Ceylan, andGozpinar (2006),this

paper claims that mood changes stemming from the

results of soccer matches affect stock market returns.

Thepresentstudyfurtherclaimsthatmoodchangesalso

affect the return–volatility relationship: agents become

morerisk averseafteralossand lessriskaverseaftera

win.

Inthisstudy,wetestwhethermoodhasasigniﬁcant

effectonassetprices.Toaccomplishthis,weﬁrstﬁndan

eventthatcanaffectthousandsofpeople’smoods

simul-taneously.Sportseventsprovidesuchanexample;soccer

matches in particular. The psychology literature

docu-mentsthatsoccerresultshave amuchlargerimpacton

mood than supposed bymost economistsbecausethey

affectmillionsofpeopleinasimilarway.Somestudies

dis-cusstheeffectof sportsresultsand mood.Forinstance,

Wann,Dolan,Mcgeorge,andAllison(1994)documentthat

fansoftenexperiencestrongpositivereactionsto

watch-ingtheirteamperformwell.Hirt,Erickson,Kennedy,and

Zillmann(1992)ﬁndthatIndianaUniversitycollege

stu-dentsestimatetheirownperformancestobesigniﬁcantly

betterafterwatchingawinbytheircollegebasketballteam

thanafterwatchingaloss.

(2)

Mainstreameconomicssupportstheeffectofsocceron

economicperformance.Pollard(2002)documentsthe

rela-tionshipbetweenthegrowthrateofEuropeancountries

andWorldCupsuccess.Moreover,Berument,Inamlik,and

Yucel(2003)andBerumentandYucel(2005)ﬁndthat

suc-cessinsocceraffectsindustrialproductioninTurkey.

Similar to the present paper, Edmans et al. (2007),

Berument andYucel (2005) andBerument etal. (2006)

examinetheeffectsofinternationalsoccermatchesona

setofmacroeconomicvariablesviamood.Edmansetal.

(2007)investigatestockmarketreactionstotheoutcome

oftheinternationalsoccermatchesof42countries, and

ﬁndthatlossesin internationalsoccermatches havean

economicallyandstatisticallysigniﬁcantnegativeeffecton

thatcountry’sstockmarket.However,thesamestudydoes

notﬁndacorrespondingeffectafterwins:returnsondays

followingwinsareclosetozeroandnotstatistically

sig-niﬁcant.Edmans etal.(2007)suggest thata soccerloss

effectiscausedbyachangeininvestormood.ForTurkey,

Berumentetal.(2006)examinetheeffectsofBesiktas’

suc-cessonstockreturnsandshowthatthisteam’swinsover

foreignrivalsincreasestockmarketreturns.Edmansetal.

(2007)andBerumentetal.(2006)considertheeffectof

lossesandwinsonlyonstockmarketreturns.Incontrast,

ourstudyconsidersnotonlythelevelofreturnbutalsothe

pricingofriskchangeswithsoccerscores.

The existing literature on behavioral ﬁnance mostly

concentratesonreturns.However,agents’riskperceptions,

whichthereturn–volatilityrelationshipmaycapture,may

alsochangewithmood.Loewenstein(2000),Romer(2000)

and Loewenstein, Weber, Hsee, and Welch(2001) note

thattheprobabilityweightingfunctionmaydependonan

agent’semotionalstate.Leahy(2002)modelsthe

invest-mentbehaviorofadepressedindividual.Accordingtothat

model,peoplewhoaredepressedbelievethattheyhave

fewpresentandfutureresourcesandbelievetheyhavea

lowutilityofgaininamarketthatisvolatileand

declin-ing.Therefore, depressedindividuals adoptstrategiesto

minimizetheirlosses.JohnsonandTversky(1983)indicate

thatmoodinﬂuencesjudgmentsofoutcomeprobability.

In one experiment, theyﬁndthat agents are less likely

togambleafteralossandmorelikelytogambleaftera

gain.Arkes,Herren,andIsen(1988)demonstratethatthe

salesof Ohio’sstate lottery ticketsincrease in thedays

afteravictorybytheOhioStateUniversitysoccerteam.

Similarly,Dohmenetal.(2006)reporttheresultsofa

tele-phone surveyconducted inGermany justone dayafter

theGermansoccerteam’sunexpectedgoodperformance

duringtheFIFA2006WorldCup.Theinterviewswere

re-performedontheafternoonofthenextday.Theauthors

notethatcurrentperceptionandfutureeconomic

expec-tationsimprovedatthepersonalandeconomy-widelevels.

HankeandKirchler(2010)arguethatastheimportance

of amatch increasesand itsresultis moreunexpected,

itsimpactwillbehigher.Edmansetal.(2007)alsonote

thatunderobjectiveprobabilities,marketdeclineisstrong

forunexpectedlosses.Thus,theeffectismorepronounced

afteralossforateamthatismorelikelytowin,i.e.,a

usu-allysuccessfulteam.Theeffectisalsomorepronounced

aftera winfora teamthat isless likelytowin.Tothis

end,wedocumentresultsfromfourcountrieswherethe

soccer is the most important sport. Teams from Spain

and the UK are considered to bemore successfulthan

teamsfromChileandTurkey.Weexpecttoseethatlosses

havelargernegativeeffectsforSpainandtheUKandwe

expectthatwinshavelargerpositiveeffectsforChileand

Turkey.

In this study, we examine the effects of the results

ofinternationalsoccermatchesagainstforeignrivalson

stock exchanges and we report the effect of wins and

lossesoffourcountries’importantsoccerteamsonstock

returns,along withtheireffects onrisk perception.We

organizetherestofthepaperasfollows:Section2

elab-oratesonthemodelthatweuseandSection3provides

the empirical evidence. The last section concludes the

paper.

2. Methodology

In order to assess the effects of soccer matches on

stockmarketreturnsandreturn–volatilityrelationships,

weuseaclassofAutoregressiveConditional

Heteroskedas-tic(ARCH)speciﬁcations.TheARCHspeciﬁcationcaptures

theconditionalvarianceofstockreturnsasameasureof

risk.

Inthispaper,weuseNelson’s(1991)EGARCH

speciﬁca-tionforconditionalmodelingbecauseofvariousappealing

features.First,itremovespartofthenon-negativity

param-eter restrictions of the conditional variance within the

traditional Bollerslev’s (1986) GARCH model. Second, it

allowsstockmarketshockstohaveasymmetriceffectson

conditionalvariance.Thatis,positiveandnegativeshocks

may affect risk perception differently, i.e., the leverage

effect.

Here,wewanttoseetheeffectsofmoodonrealstock

returns and return–volatility relationships. Our model

speciﬁcationis: Rt=xtˇ+ht+·outcomet+ϕ·outcomet·ht+εt (1) εt∼(0,h2t) logh2 t =+ p

i=1 ıilogh2t−i+ q

j=1 j

εt−j ht−j

−E

εt−j ht−j

+εt−j ht−j

(2)

whereRt isthestockreturncalculatedbytakingthe

log-arithmicﬁrstdifferenceofthedailyclosingpriceofstock

marketsandx

tisavectorforexogenousvariablesattime

t.ThemodelincludesdailydummiesforMonday,Tuesday,

Wednesday,ThursdayandFridaytoaccountforthe

day-of-the-weekeffect.Thexvectoralsoincludeslaggedvaluesof

thereturnbecauseCosimanoandJansen(1988)arguethat

iferrorsareautocorrelated,thenEngle’s(1982)ARCH-LM

testsuggeststhepresenceoftheARCHeffectevenwhenit

isnotpresent.Theconditionalstandard deviationofthe

residualterm,ht,is taken asa proxyfor risk attime t;

outcometisthedummyvariableforthewinsorlossesand

takesthevalueofonefor thenextbusinessdayaftera

winoralossininternationalmatches,andzerootherwise.1

1 _We_note_that_within_the_sample_period_there_were_never_two_or_more

wins(orlosses)tobeconsideredonthenextbusinessday.Therefore, outcometnevertakesthevalueoftwoorabove.

(3)

Outcomet·htistheinteractivetermforwinsorloseswith

theconditionalstandarddeviationoftheresidualterm.The

interactivetermisintendedtocapturethechangeinthe

return–volatilityrelationshipwithwinsorlossesandεtis

theerrortermattimet.Asweexpectapositive

relation-shipbetweenwinsandreturns,wealsoexpectanegative

relationshipbetweenlossesandreturns.Thisexpectation

impliesthatthesignoftheestimatedoutcometcoefﬁcient,

,willbepositiveforthewinsofChileandTurkeyand

neg-ativeforthelossesofSpainandtheUK.Ontheotherhand,

thetheoryonchangeinriskaversionsuggeststhatthesign

ofthecoefﬁcientoftheinteractivetermbetweenht and

outcomet,ϕ,willbenegativeforChileandTurkeyand

pos-itiveforSpainandtheUK.Thispredictionisbecausewins

makepeoplelessriskaverse,ormoreriskloving,andlosses

makepeoplemoreriskaverse,orlessriskloving;thuswe

expecttoacceptthesamereturnwithahigherriskora

lowerreturnandalowerriskorahigherreturnwiththe

samerisklevel.

In Eq. (2), the parameter captures residual

asym-metric effects. When =0, a negative surprise, that is,

(εt−1/ht−1)<0, has the same effect on volatility as a

positivesurpriseof equalmagnitude.When−1<<0,a

negativesurpriseincreasesvolatilitymorethanapositive

surprisedoes.When<−1,apositivesurprisedecreases

volatilityandanegativesurpriseincreasesvolatility.

Teamsgenerallyplay alltheirmatchesafterbusiness

hours.MostmatchesareonTuesdays,Wednesdaysand

Thursdays.Therefore,thedummyvariable forwins and

losses,outcomet, forthe nextbusinessdaycan capture

higherreturnsassociatedwiththesedays.However,itis

importanttorecognizedummyvariablesforthe

day-of-the-weekeffect;thus,estimatesforwinsandlossescapture

theeffects ofwinsand lossesaftertheday-of-the-week

effectisaccountedfor.

3. Empiricalevidence

Toassesstheeffectof internationalsoccer scoreson

the stock market, we gather data from four countries

where soccer is an important sports activity;

interna-tionalordomesticsuccessesbelongtoafew teamsthat

havesupportersalloverthecountry.Mostpeople

asso-ciatethemselveswithoneofthefewteams,whichfurther

ensuresthatthesoccerresultswillhaveacountry-wide

effect.Forthesereasons,wechooseChile,Spain,Turkey

and theUK. We select Cobreloa, Colo Colo, Universitad

CatolicaandUniversitaddeChilesoccerteamsforChile;

BarcelonaandRealMadridsoccerteamsforSpain;

Besik-tas,FenerbahceandGalatasaraysoccerteamsforTurkey

and Arsenal, Chelsea, Liverpool and Manchester United

soccer teams for the UK. The time span of the data is

from 01 January 1985 to 02 February 2007, but may

varyamongcountries. Wegatherthedataforthestock

exchangesfromDatastream.Wecomputethestock

mar-ketreturnsbytakingthepercentagechangeofthedaily

closingpriceoftheGeneralStockPriceIndex(IGPA)for

Chile,theMadridStockExchangeGeneralIndexforSpain,

theIstanbulStockExchange100IndexforTurkey,andthe

FTSE100Index fortheUK.We obtaintheinternational

soccermatchresultsforthesefourcountriesmainlyfrom

http://www.rsssf.com.2

Table 1 reports the time period, teams, and

num-ber of international wins, losses, and ties, along with

theirpercentagestototalmatches.Table1suggeststhat

in terms of results, Spain and the UK have more

suc-cessful soccer teams than Chile and Turkey. Because

of the differences in results, we might observe

differ-enteffects oflossesand winsbetweenthetwocountry

sets.

Table2reportsdescriptivestatisticsforthestock

mar-ketreturnsofeachcountry.Notethatexcesskurtosis,the

kurtosisvalueisgreaterthanthree, andthestatistically

signiﬁcantJarque-BeraTestsfornormalitysuggest

time-varyingvariancesforthereturns.

Table3reportstheestimatesofsoccermatchoutcomest

on stock market returns and return–volatility

relation-ships for the four countries. The estimated coefﬁcients

of outcomet and outcomet·ht are important; however,

we discuss other coefﬁcients to show that our

ﬁnd-ings parallel the existing literature on the respective

models.

First, the Mondayeffect is negative for allcountries

but Spain,and Tuesday has a negative and statistically

signiﬁcant effect for Spain.Wednesday alsohas a

neg-ative and statistically signiﬁcant effect for Spain only.

Thursdays’coefﬁcientsarepositiveforallbutSpain.For

Chile,theThursdayeffectispositiveandstatistically

sig-niﬁcant. Friday effects are positive and are statistically

signiﬁcantfor Chileat 1%,for Turkeyat5% and forthe

UKat10%signiﬁcancelevels.Therefore,Mondayreturns

arelowerthanFridayreturnsforallcountriesbutSpain.

However,FridayreturnsarehigherthanTuesdayreturns

for Spain. Thus, our ﬁndings are parallel to the

ﬁnd-ingsintheliteratureontheday-of-the-weekorweekend

effects.

Second,therisktermcoefﬁcient,ht,ispositiveforall

countries. The effectis statisticallysigniﬁcantat the1%

signiﬁcance levelfor Spain,and fortheUKit is

statisti-callysigniﬁcantatthe10%signiﬁcancelevel.Thepositive

coefﬁcientssuggestthatriskincreasesexpectedstock

mar-ket returns.Thisﬁndingis inlinewiththemainstream

asset-pricing model, which indicates the association of

higherreturnswithriskierportfolios.

Third,consistentwiththeexistingliterature,outcomet

coefﬁcientsarenegativewhentheSpanishandBritish

soc-certeamsloseandpositivewhentheChileanandTurkish

soccer teamswin. Thisﬁndingsupports theproposition

thathighermoraleincreasesstockmarketreturnsforthe

cases of wins and decreasesreturns after a loss. These

resultsmakesensebecauseSpainandtheUKare

success-fulcountriesinsoccer.Thus,winsmaynotaffectthemood

ofsoccerfansasmuchaslosses.Forthisreason,weexpect

theeffectsoflossestobehigherforthosecountries.Onthe

otherhand,ChileandTurkeyarenotassuccessfulinsoccer

asSpainandtheUK.Thus,winsaremorelikelythanlosses

2_We_also_use_{http://www.statto.com/}_,_{http://kassiesa.home.xs4all.nl/} bert/uefa/data/index.html and http://www.eurocupshistory.com/uefa cup/toﬁndorvalidatesomeofthematchresults.

(4)

Table1

Thenumberofwins,tiesandlossesforeachcountry.

Sampleperiod Numberscoreson

Wins Ties Losses

Chile 02January1987–31December2006 94 73 90

(37%) (28%) (35%) Cobreloa 22 18 26 (33%) (27%) (40%) ColoColo 31 22 25 (40%) (28%) (32%) UniversitadCatolica 30 24 25 (38%) (30%) (32%) UniversitaddeChile 11 9 14

Spain 01November1977–02February2007 298 112 127

(55%) (21%) (24%)

Barcelona 151 62 60

(55%) (23%) (22%)

RealMadrid 147 50 67

(56%) (19%) (25%)

Turkey 03July1987–02February2007 99 65 109

(36%) (24%) (40%) Besiktas 27 19 30 (36%) (25%) (39%) Fenerbahce 24 11 35 (34%) (16%) (50%) Galatasaray 48 35 44 (38%) (28%) (34%)

UnitedKingdom 01January1985–02February2007 249 120 107

(52%) (25%) (23%) Arsenal 59 33 31 (48%) (27%) (25%) Chelsea 47 17 20 (56%) (20%) (24%) Liverpool 63 29 25 (54%) (25%) (21%) ManchesterUnited 80 41 31 (53%) (27%) (20%)

Source:Thedataonthesoccerscoresarefromhttp://www.rsssf.com.

Numbersinparenthesesarethepercentagesofthecorrespondingoutcomes(wins,tiesorlosses)tototalmatches.

toaffectthestockmarketreturnandthereturn–volatility

relationship.

Finally,weconsidertheinteractiveterm,outcomet·ht.

ThecoefﬁcientsarenegativeforChileandTurkeyand

posi-tiveforSpainandtheUK.ThenegativecoefﬁcientsforChile

andTurkeysuggestthatmarketscompensateagentsless

afterawinforbearingthesamelevelofrisk;thatis,agents

aremorerisklovingafterawin.Thepositivecoefﬁcientsfor

SpainandtheUKsuggestthatmarketscompensateagents

moreafteralossforbearingthesamelevelofrisk;agents

becomemoreriskaverse.Weestimatethesame

speciﬁca-tionsforSpain’sandtheUK’swinsandChile’sandTurkey’s

losses.Wedonotreporttheseestimatesheretosavespace

buttheyareavailablefromtheauthorsuponrequest.

Nei-therthe outcome coefﬁcients northe interactiveterms

are statistically signiﬁcant, supporting the proposition

Table2

Descriptivestatisticsofthecountries’stockmarketreturnsintherespectivesampleperiods.

Chile Spain Turkey UnitedKingdom

Mean 0.075580 0.050450 0.217808 0.032523

Standarddeviation 0.842854 1.057097 2.987632 1.006494

Skewness −0.258825 −0.118161 0.180407 −0.586982

Kurtosis 11.659901 5.622417 3.028642 9.673965

Jarque-Bera 29457.843566 10020.099049 1837.702786 22676.560464

Source:DataforthestockexchangesarefromDatastream.

Notes:Returnsarecalculatedbytakingthepercentagechangeofthedailyclosingpriceofthestockexchangeindexes.WeusedtheGeneralStockPrice Index(IGPA)forChile,theMadridStockExchangeGeneralIndexforSpain,theIstanbulStockExchange100IndexforTurkey,andtheFinancialTimesStock Exchange100IndexfortheUK.

(5)

Table3

Effectsofsoccermatchoutcomesonstockreturnsandthereturn–volatilityrelationship.†

Outcome– losses Outcome–wins

Spain UK Chile Turkey

PanelA:estimatesforthereturnequation

Monday 0.0390* _−0.0145 _−0.1268*** _−0.1728 (0.0761) (0.6509) (0.0000) (0.1122) Tuesday −0.0814*** _0.0303 _0.0092 _−0.1428 (0.0000) (0.4433) (0.7355) (0.1997) Wednesday −0.0485* _0.0425 _0.0290 _0.0744 (0.0643) (0.1824) (0.2944) (0.5199) Thursday −0.0107 0.0223 0.0671** _0.0774 (0.6087) (0.4859) (0.0155) (0.4917) Friday 0.0224 0.0732* _0.0782*** _0.2383** (0.5070) (0.05384) (0.0051) (0.0370) Outcomet −0.1298** −0.5265*** 0.3110** 1.2969** (0.0306) (0.0007) (0.0421) (0.0423) ht 0.0438*** 0.0139* 0.0246 0.0457 (0.0000) (0.0815) (0.5294) (0.2550) Outcomet·ht 0.0969* 0.6350*** −0.6318** −0.6102** (0.0571) (0.0002) (0.0233) (0.0330) PanelB:estimatesfortheEGARCHspeciﬁcations

Parameters K 0.0102*** _−0.0048 _−0.0212** _0.1027*** (0.0051) (0.2605) (0.0120) (0.0000) logh2 t−1 0.9893*** 0.1230*** 0.5426*** 1.0198*** (0.0000) (0.0000) (0.0000) (0.0000) logh2 t−2 0.8470*** 0.4028*** −0.3890** (0.0000) (0.0000) (0.0148) logh2 t−3 0.3224*** (0.0002)

εt−1 ht−1

−E

εt−1 ht−1

+ εt−1 ht−1

0.3015*** _0.1477*** _0.4011*** _0.3638*** (0.0000) (0.0000) (0.0000) (0.0000)

εt−2 ht−2

−E

εt−2 ht−2

+ εt−2 ht−2

−0.0746** _0.1237*** (0.0204) (0.0000)

εt−3 ht−3

−E

εt−3 ht−3

+ εt−3 ht−3

−0.0510* _−0.0200 (0.0752) (0.4829) −0.0939** _−0.3877*** _0.0211 _−0.0764** (0.0152) (0.0000) (0.5730) (0.0387) TT 4.8537*** _11.1588*** _5.6742*** _7.4441*** (0.0000) (0.0000) (0.0000) (0.0000) Functionvalue −5155.5998 −3952.5773 −2176.8415 −8510.7244 PanelC:Ljung-BoxQ-statistics

Lags

5 0.1118 0.5300 0.0297** _0.0195**

10 0.3229 0.9171 0.1784 0.1531

20 0.0859* _0.8559 _0.1642 _0.5242

60 0.4049 0.8722 0.2672 0.8138

PanelD:ARCH-LMtests Lags

5 0.9974 0.0016*** _0.0383** _0.4072

10 1.0000 0.0331** _0.1909 _0.0152**

20 1.0000 0.1747 0.5037 0.0584*

60 1.0000 0.8863 0.4444 0.2204

PanelE:non-parametric(sign)tests

Signbiastest 0.3570 0.7859 0.0298** _0.3159

Negativesizebiastest 0.4256 0.2170 0.4949 0.4659 Positivesizebiastest 0.4521 0.2515 0.0812* _0.3155

Jointtest 0.3625 0.3856 0.1546 0.3595

Source:Thedataonsoccerscoresarefromhttp://www.rsssf.comandthedataonreturnsfromDatastream.

† _We_report_p-values_in_parentheses._The_estimated_{coefﬁcients}_for_the_lagged_values_of_the_dependent_variables_have_not_been_reported_to_save_space.

***_Indicates_the_level_of_{significance}_at_the_1%_level. ** _Indicates_the_level_of_{significance}_at_the_5%_level. * _Indicates_the_level_of_{significance}_at_the_10%_level.

that losses have more of an effect on the countries

withmoresuccessfulsoccerteams;however,winshave

more of an effect on countries with less successful

teams.

The estimates for the EGARCH model and a

bat-tery of speciﬁcation tests for the model are reported

in Table 3’s Panels B–E. We show the estimated

(6)

the countries under consideration, the coefﬁcients of

log(_h_t−i)2 and

_ε_t−j/_h_t−j

₋_E

_ε_t−j/_h_t−j

₊_ε_t−j/_h_t−j

are

signiﬁcantlydifferentfromzero.3 _The_lagged-value

esti-matedcoefﬁcientsoflog(h2

t)arestatisticallysigniﬁcantand

thecharacteristicrootsofthepolynomialareallinsidethe

unitcircle.Thus,noneoftheconditionalvariance

speciﬁ-cationsisexplosive.Evidenceontheleverageeffect,the

estimated coefﬁcientfor theleverage effect,, is

nega-tiveforSpain,TurkeyandtheUK,indicatingthatnegative

surprisesincreasevolatilitymorethanpositivesurprises.

However,wedonotobserveastatisticallysigniﬁcant

lever-ageeffectforChile.

TosetupthemaximumlikelihoodestimatorforEqs.

(1)and (2),weassumethaterrors haveaGeneralError

Distribution.TTistheparameterforreturntailthickness.

ATT>2impliesthaterrorshaveathickerdistributionthan

normal.

Panel C reports the p-values of the Ljung-Box

Q-statistics forthestandardized squaredresiduals, (ε2

t/h2t),

calculatedtotestthenullhypothesisofzero

autocorrela-tionupto60lags.Overall,wecannotrejectthenullofno

autocorrelationforanycountry.Theonlyexceptionsare

Spainat20lagswithasigniﬁcancelevelat10%,andChile

andTurkeyatﬁvelagswithasigniﬁcancelevelat5%.

Next, totestfor thenull hypothesisthat there isno

ARCHeffectforthestandardizedresiduals(εt/ht),weapply

Engle’s(1982)ARCH-LMtest.Teststatisticsforthe

stan-dardizedresidualsarereportedinPanelC.Totestthenull

hypothesisthatthereisnoARCHeffect,weregressedthe

squaredstandardizedresidualsonaconstanttermandon

itsﬁfth,tenth,twentiethandsixtiethlags.Here,theaimis

totestwhetherlagtermsarejointlystatisticallysigniﬁcant;

theresultsshowthatthenullofnoARCHeffectisrejected

onlyforChileatﬁvelagsatthe5%signiﬁcancelevel,for

Turkeyat10lagsatthe5%signiﬁcanceleveland20lagsat

the10%signiﬁcancelevelandfortheUKatﬁveand10lags

atthe1%and5%signiﬁcancelevels,respectively.

PanelEreportsthenon-parametricsignandsizebias

testsperformedforthestandardizedresiduals(εt/ht).

Cal-culatingtheteststatistics,weusestandardizedresiduals

(εt/ht).We addtwo dummyvariables,mt andpt,tothe

equationsuchthatmt=1ifthenormalizedresidualis

neg-ative,0otherwise,andpt=1ifitispositive,0otherwise.

Wealsodeﬁnetwointeractivedummyvariables:smtand

spt.

smt=pt∗(εt/ht) and spt=pt∗(εt/ht)

smt=pt·(εt/ht) and spt=pt·(εt/ht)

Later,weregress(εt/ht)ontheconstantterm,mt,smtand

spt.Inthesigntest,wedeterminewhetherthecoefﬁcient

ofmiszeroornot.Withthenegativesigntest,we test

whetherthecoefﬁcientofsmt iszero;withthepositive

signtestwetestwhetherthecoefﬁcientofspiszeroand

withthejointtestwetestthesenullhypothesesjointly.

3_We_determine_the_selection_of_the_lag_order_for_both_terms_such_that

thespeciﬁcationtestsreportedinPanelsB,C,DandEdonotrevealthat standardizedresidualsarenotiid.

Theresultsshowthatnocountry’steststatisticsare

statisti-callysigniﬁcant,exceptforthesignbiasedtestandpositive

sizebiastestforChile,whicharestatisticallysigniﬁcantat

the5%and10%levels,respectively.Hence,sincemosttest

statisticsreportedinPanelsB–Earenotstatistically

signif-icant,weconcludethattheysupportourspeciﬁcations.

4. Conclusions

This paper presents empirical evidence that soccer

matchscoresaffectstockmarketreturnsandstockmarket

return–volatility relationships. Evidence from countries

withrelativelymoresuccessfulsoccer teams,Spainand

theUK,indicate stockmarket returnsdecrease and the

risk–returnrelationshipchangeswiththematchscoreso

thatagentsbecomemoreriskaverseafteraloss.However,

we cannotﬁnd statisticallysigniﬁcantevidencefor this

changeafterawin.Thedatafromcountrieswithrelatively

lesssuccessfulsoccerteams,ChileandTurkey,revealthat

stockmarketreturnsincreaseandagentsbecomemorerisk

lovingafterawin.Similarly,wecannotﬁndstatistical

sig-niﬁcanceafterlosses.Themore-pronouncedlosseffectfor

SpainandtheUKandwineffectforChileandTurkeymaybe

duetosuccessfulhistoriesorfanexpectationsfromthese

countries’soccerteams.

References

Arkes,H.,Herren,L.,&Isen,A.(1988).Theroleofpotentiallossinthe inﬂuenceofaffectonrisk-takingbehavior.OrganizationalBehaviorand HumanDecisionProcesses,66,228–236.

Berument,H.,Ceylan,N.B.,&Gozpinar,E.(2006).Performanceofsoccer onthestockmarket:EvidencefromTurkey.SocialScienceJournal,43, 695–699.

Berument,H.,Inamlik,A.,&Yucel,E.(2003).Theeffectoffootballon productivity.IktisatIsletmeveFinans,212,51–62.

Berument,H.,&Yucel,E.(2005).LongliveFenerbahce:Production boost-ingeffectsofsoccerinTurkey.JournalofEconomicPsychology,26, 842–861.

Bollerslev, T. (1986). Generalized autoregressive conditional het-eroskedasticity.JournalofEconometrics,31,307–327.

Cosimano,T.F.,&Jansen,D.W.(1988).EstimatesofthevarianceofU.S. inﬂationbasedupontheARCHModel.JournalofMoney,Credit& Bank-ing,20(3),409–421.

Dohmen,T.,Falk,A.,Huffman,D.,&Sunde,U.(2006).Seeminglyirrelevant eventsaffecteconomicperceptionsandexpectations:TheFIFAWorldCup 2006asanaturalexperiment.IZADiscussionPapersNo:2275. Edmans,A.,Garcia,D.,&Norli,O.(2007).Sportssentimentandstock

returns.JournalofFinance,62,1967–1998.

Engle,R.F.(1982).Autoregressiveconditionalheteroskedasticitywith estimatesofthevarianceofU.K.inﬂation.Econometrica,50,987–1008. Hanke,M., & Kirchler, M.(2010). Football championships and Jersey sponsors’stockprices:Anempiricalinvestigation,WorkingPapersin EconomicsandStatistics,UniversityofInnsbruck.

Hirt,E.R.,Erickson,G.A.,Kennedy,C.,&Zillmann,D.(1992).Costsand ben-eﬁtsofallegiance:Changesinfans’self-ascribedcompetenciesafter teamvictoryversusdefeat.JournalofPersonalityandSocialPsychology, 63,724–738.

Johnson,E.J.,&Tversky,A.(1983).Affect,generalization,andthe percep-tionofrisk.JournalofPersonalityandSocialPsychology,45,20–31. Leahy,R.L.(2002).Psychologyandtheeconomicmind:Cognitiveprocesses

&conceptualization.NewYork:SpringerPublishingCompany. Loewenstein,G. (2000).Emotionsineconomic theoryandeconomic

behavior.AmericanEconomicReview,90(2),426–432.

Loewenstein,G.F.,Weber,E.U.,Hsee,C.K.,&Welch,N.(2001).Riskas feelings.PsychologicalBulletin,127(2),267–286.

Nelson,D.(1991).Conditionalheteroskedasticityinassetreturns:Anew approach.Econometrica,59,347–370.

Pollard,P.S.(2002).Grooowwwth!.InternationalEconomicTrends,Annual Edition.July1-1.

(7)

Romer,P.M.(2000).Thinkingandfeeling.AmericanEconomicReview, 90(2),439–443.

Stracca,L.(2004).Behavioralﬁnanceandassetprices:Wheredowestand? JournalofEconomicPsychology,25(3),373–405.

Wann,D.L.,Dolan,T.J.,Mcgeorge,K.K.,&Allison,J.A.(1994). Relation-shipsbetweenspectatoridentiﬁcationandspectators’perceptionsof inﬂuence,spectators’emotions,andcompetitionoutcome.Journalof Sport&ExercisePsychology,16,347–364.