WORKI NG PAPER SERI ES

NO 1424 / FEBRUARY 2012

THE PITCH RATHER THAN THE PIT
INVESTOR INATTENTION DURING
FIFA WORLD CUP MATCHES
by Michael Ehrmann and David-Jan Jansen
NOTE: This Working Paper should not be reported as representing the
views of the European Central Bank (ECB). The views expressed are
those of the authors and do not necessarily re ect those of the ECB.
In 2012 all ECB
publications
feature a motif
taken from
the 50 banknote.
European Central Bank, 2012
Address
Kaiserstrasse 29, 60311 Frankfurt am Main, Germany
Postal address
Postfach 16 03 19, 60066 Frankfurt am Main, Germany
Telephone
+49 69 1344 0
Internet
http://www.ecb.europa.eu
Fax
+49 69 1344 6000
All rights reserved.
ISSN 1725-2806 (online)
Any reproduction, publication and reprint in the form of a different publication, whether printed or produced electronically, in whole
or in part, is permitted only with the explicit written authorisation of the ECB or the authors.
This paper can be downloaded without charge from http://www.ecb.europa.eu or from the Social Science Research Network electronic
library at http://ssrn.com/abstract_id=2004234.
Information on all of the papers published in the ECB Working Paper Series can be found on the ECBs website, http://www.ecb.europa.
eu/pub/scientic/wps/date/html/index.en.html
Acknowledgements
This paper presents the authors personal opinions and does not necessarily reect the views of the European Central Bank, de Neder-
landsche Bank or the Eurosystem. The rst author thanks de Nederlandsche Bank for its hospitality during his stay as visiting scholar.
We thank Thomas Kostka for help in retrieving the data, and Yener Altunbas, Boppo J anssen, Luis Marques, Marco Massarenti and
seminar participants at various venues for useful comments and discussions.
Michael Ehrmann (Corresponding author)
at European Central Bank, Kaiserstrasse 29, D-60311 Frankfurt am Main, Germany; e-mail: [email protected]
David-Jan Jansen
at De Nederlandsche Bank, Westeinde 1, PO Box 98, 1000 AB Amsterdam, the Netherlands; e-mail: [email protected]

1
Abstract
At the 2010 FIFA World Cup in South Africa, many soccer matches were played during
stock market trading hours, providing us with a natural experiment to analyze fluctuations in
investor attention. Using minutebyminute trading data for fifteen international stock
exchanges,wepresentthreekeyfindings.First,whenthenationalteamwasplaying,thenumber
of trades dropped by 45%, while volumes were 55% lower. Second, market activity was
influencedbymatchevents.Forinstance,agoalcausedanadditionaldropintradingactivityby
5%.Themagnitudeofthisreductionresembleswhatisobservedduringlunchtime,andassuch
mightnotbeindicativeforshiftsinattention.However,ourthirdfindingisthatthecomovement
between national and global stock market returns decreased by over 20% during World Cup
matches, whereas no comparable decoupling can be found during lunchtime. We conclude that
stock markets were following developments on the soccer pitch rather than in the trading pit,
leadingtoachangedpriceformationprocess.

JELcodes:G12,G14,G15
Keywords:investorinattention,stockmarkets,tradingvolume,highfrequencydata,soccer

2
Nontechnicalsummary
Every fouryears,32nationalsoccerteamscompeteintheWorldCup.Thistournament,
whichisorganisedbytheworldsoccerassociationFIFA,attractsattentionfrommillionsoffans
across the globe. During the 2010 edition in South Africa, many matches were played during
stock market trading hours. This presents us with a natural experiment to analyze possible
fluctuationsininvestorattention.
The paper presents three key findings. First, we find strong evidence of decreased
activityinstockmarketsduringsoccermatchesatthe2010WorldCup.Tradingactivitydropped
markedly, especially if the national team was one of the competitors. Compared to normal
market circumstances, the median number of trades dropped by 45% if the national team was
playing,whilethevolumedroppedbyaround55%.Second,weshowhowgoalsscoredbyeither
team led to an even stronger decline in the number of trades and offered quotes. Also, we find
that market activity was already significantly below the benchmark right before the match
started,andcontinuedtobelowerduringthe45 minutes after thematch hadended.Third,we
show that also price formation was affected during the soccer matches, as the evolution of
returnsonnationalmarketsdecoupledfromthoseonglobalmarkets.
Overall, there is a strong sense that stock markets were following developments on the
soccerpitchratherthaninthetradingpit.Theseresultsprovideevidenceforlimitedattentionin
financial markets, which in itself affects the price formation process. Further tests show that
inattention was particularly strong for relatively less salient information there was a
particularly strong decoupling of national from global markets as long as the price movements
on the global market were relatively small. Furthermore, the crosssectional dispersion of
returns across the individual constituents of a countrys stock market index was substantially
reduced,suggestingthatthedistractioncomingfromthesoccermatchesledtoareducedfocus
onfirmspecificasopposedtomarketandsectorwideinformation.

3
1.Introduction
Everyfouryears,32nationalsoccerteamscompeteintheWorldCup.Thistournament,whichis
organised by the world soccer association FIFA, attracts attention from millions of fans across
the globe. During the 2010 edition in South Africa, many matches were played during stock
market trading hours. This presents us with a natural experiment to analyze fluctuations in
investor attention. We show that during matches at the 2010 World Cup, trading activity on
fifteeninternationalstockexchangesdeclinedsharply,especiallyifthenationalteamwasoneof
the contenders. Furthermore, price formation on national markets decoupled from global
markets.Overall,thereisastrongsensethatstockmarketswerefollowingdevelopmentsonthe
soccerpitchratherthaninthetradingpit.
Thispaperfitsintheliteratureonlimitedattentioninfinancialmarkets,whichtakesits
cue from the idea that attention is a scarce resource (Kahnemann 1973).
1
In recent years,
supportive empirical evidence has accumulated steadily. For instance, Cohen and Frazzini
(2008) present evidence that news about a given firm is immediately reflected in that firms
stock price, but only affects stock prices of economically related firms with some delay,
suggesting that there is inattention to relatively complex information. Peng and Xiong (2006)
show that investors process more market and sectorwide information than firmspecific
information, implying they are inattentive to relatively detailed information. Furthermore,
DellaVignaandPollet(2007)suggestthatstockmarketvaluationsofagesensitivesectorstocks
(suchastoys,beer,ornursinghomes)neglect(publiclyavailable)demographicinformation,as
investorsareinattentivetoinformationaboutthedistantfuture.
Furthermore, Mondria et al. (2010) show how attention allocation across countries can
helptoexplainhomebias,whileHou,PengandXion(2009)focusondifferentialattentionacross
firms. They show that less attention given to a stock leads to a muted price reaction to that
companys earnings announcements. Barber and Odean (2008) find that individual investors
tend to be net buyers of attentiongrabbing stocks, either meaning those that are discussed in
the news, that experience high abnormal trading volume, or extreme oneday returns. This
findingalsosuggeststhatsalientinformationreceivesmoreattention.
Finally, a set of papers analyses variation in allocation of attention across time:
DellaVigna and Pollet (2009) have argued that the upcoming weekend distracts investors and
mutes the response to news released on a Friday. Indeed, they find that earnings
announcementsreleasedonaFridayhaveafifteenpercentlowerimmediateresponseanda70

1
A related literature deals with inattention in macroeconomics, see, e.g., Mackowiak and Wiederholt
(2009),MankiwandReis(2002)orSims(2003).

4
percent higher delayed response. Also, trading volume is significantly lower around Friday
announcements. Louis and Sun (2010) have shown how market reactions have also been less
strongwhenmergerannouncementsweremadepubliconaFriday.Finally,Hirshleifer,Limand
Teoh (2009) have found that price and volume reactions to earnings surprises are weaker on
dayswhenagreaternumberofotherfirmsreleasesimilarinformation.
2

A major challenge in this literature is how to identify variation in investor attention.

Corwin and Coughenour (2008) propose a direct test by studying market makers on the New
York Stock Exchange, who often have to divide attention across a range of securities. They
measurethedegreeofattentionaspecialistcangiveanindividualstockbytheinverseoftrading
volume and absolute returns of all stocks in her portfolio. Our paper proposes an alternative
measure for shifts in investor attention across time: major sporting events. The 2010 FIFA
World Cup is particularly suited for investigating such an exogenous shift, as a substantial
amountofsoccermatcheswereplayedduringtradinghoursinEurope,theUnitedStates,Latin
America and South Africa. Another distinguishing feature of our paper is the use of minuteby
minutedata onstockmarketactivity.Usingdata onthenumber oftrades andtrading volumes,
we can accurately pinpoint shifts in attention. In particular, we are able to relate attention to
matchevents,suchasthescoringofgoalsorthehalftimebreak.
Wearenotthefirsttostudytheeffectsofsportingeventsonfinancialmarkets.Edmans,
GarcaandNorli(2007)haveshownthatlosinganinternationalsoccermatchcanleadtolower
nextdayreturnsonthenationalstockmarketofupto49basispoints.Forinternationalcricket,
rugby and basketball games, the loss effect is smaller, but still significant. Kaplanski and Levy
(2010)arguedthatthelosseffectfoundbyEdmansetal.canevenaffecttheU.S.stockmarket,as
alargeshareofinvestorsinmostcountriesinvestinternationally.
3

Major sporting events seem a likely candidate to study shifts in attention, as so many
people follow these tournaments. During the 2010 World Cup more than three million
spectatorsattendedthe64soccermatches.
4
However,mostfansfollowedmatchesontelevision,
either at home or in public places. The final match between Spain and the Netherlands was

2
Other contributions include Gabaix, Laibson, Moloche and Weinberg (2006) and Hirshleifer and Teoh
(2003).
3
Another debate focuses on net economic benefits of hosting major sporting events. As it turns out, it is
difficulttoestablishclearpositivefallout.BaadeandMatheson(2004)estimatethatU.S.hostcitiesofthe
1994WorldCupexperiencedcumulativelossesof$5.5to9.3billion.Takingalongrunperspective,Hagn
andMaennig(2008)findthatthatthe1974WorldCupinWestGermanywasunabletogeneratemedium
tolongrunpositiveeffectsonemployment.RoseandSpiegel(2011)provideamorepositiveassessment,
byshowinghowexportsofcountrieshostingtheOlympicGameshavebeenpositivelyaffected.
4
http://www.fifa.com/worldcup/archive/southafrica2010/statistics/news/newsid=1273493/index.html

5
watchedby700millionviewers,accordingtoearlyestimatesbyFIFA.
5
WhentheWorldCupwas
held in Germany in 2006, 376 channels broadcasted the matches, and the total cumulative
television audience for all matches during the 2006 World Cup was an amazing 26.3 billion.
Furthermore, interest in the soccer matches is not restricted to one particular region. In 2006,
Asiawasthebiggestcontributortothenumberofviewers,withmorethaneightbillioninhome
viewers during all matches, while audiences were also large in traditional soccer regions like
LatinAmericaandEurope.IntheUnitedStates,audiencesincreasedbyaround40%compared
to the 2002 edition.
6
According to Nielsen, 112 million people in the US watched part of the
gamesofthe2010WorldCup.
7

When major sporting events take place during business hours, are people tempted to
watch? Lozano (2011) estimates that the average worker in the United States reduced his
working hours during the World Cups from 1994 to 2006 by between 9 and 28 minutes per
week.Insteadoftakingtimeoff,workerscanalsocallinsick.Despitethisbeingamajorconcern
for employers, in a 2010 survey only 3% of workers in the United Kingdom said they would
considertakingtimeoffsicktowatchgames.
8
Ontheotherhand,SkogmanThoursie(2004)has
estimated that the number of Swedish men who reported sick increased in parallel to major,
televisedsportingevents,suchastheOlympicGames.
Is it possible that financial markets could also be distracted by sporting events? Would
this not contrast the notion of continuously operating markets, where information is tirelessly
processed 24 hours per day? Perhaps, but the literature on limited attention cited above
stronglysuggeststhatthisviewoffinancialmarketsmaybeoptimistic.Inresponsetoupcoming
soccer matches, investors may place orders well in advance of the matches, or, alternatively,
postponethemtothenextday.Alternatively,investorsmaystillbepresentattheirdeskduring
matches,butmaybedistractedastheactionunfolds.Intheend,thisquestioncanonlybesettled
empirically. Therefore, we assembled minutebyminute data for stock markets of fifteen
countries that participated in the 2010 World Cup. We gathered data on the number of ticks,
trading volumes and offered quotes for all stocks included in a countrys main stock index. We
alsocollectedinformationonthepricehistoryforthenationalstockindex,aswellasindividual
stockprices.
Using our highfrequency dataset, we present three key findings. First, we find strong
evidence of decreased activity in stock markets during soccer matches at the 2010 World Cup.

5
http://www.reuters.com/article/2010/07/13/usfootballidUSTRE66C0ZV20100713
6
http://www.fifa.com/aboutfifa/marketing/factsfigures/tvdata.html
7
http://blog.nielsen.com/nielsenwire/consumer/increasedviewershiponlinevisitsandad
engagementamongworldcuphighlights/.
8
http://www.kronos.co.uk/PR/AbsenceSurvey2010.aspx

6
Trading activity drops markedly, especially if the national team is one of the competitors.
Compared to normal market circumstances, the median number of trades drops by 45% if the
national team is playing, while the volume drops by around 55%. Even though there is a clear
heterogeneity in the magnitude of this effect across countries, we can identify a reduction in
everysingleoneofthem.Thedeclineintradingactivityisalsoaccompaniedbyalowerlevelof
activity by market makers. During national team matches, the number of offered quotes is
roughly30%lowerthanotherwise.
Second, exploiting the minutebyminute frequency of our data, we show how goals
scored by either team led to an even stronger decline in the number of trades and offered
quotes. Also, we find that market activity is already significantly below the benchmark right
before the match starts, and continues to be lower during the 45 minutes after the match has
ended. During the halftime break, trading activity recovers somewhat compared to the actual
playingtime,butremainssubstantiallylowerthanonbenchmarkdays.
Third,weshowthatalsopriceformationisaffectedduringthesoccermatches.Assuming
thatthepriceformationonglobalstockmarketsonaverageshouldbechangingonlymarginally,
whereas national stock markets are more likely to be affected, we test whether and how the
comovement between national and global stock market returns change during matches of the
national team. On average, this comovement is found to be reduced by somewhat more than
20%,anditissmallerbymorethan40%ifreturnsintheglobalmarketarerelativelysmall(and
thereforelesssalient).Importantly,duringlunchtime,whentradingactivitydeclinesbyasimilar
orderofmagnitude,thereisonlylimitedevidenceofdecouplingfromglobalstockmarkets.This
comparisonsuggeststhatsoccergeneratedrelativelyhighlevelsofinattention,whichcoincided
withlessfrequentupdatingofstockprices.Inlinewiththis,wealsofindasignificantdeclineof
around 20% in the dispersion of individual stock returns, which should be expected if markets
price only relatively salient information (such as global news), and neglect more idiosyncratic
information(suchasforinstancestockspecificnews).
Thispaperproceedsasfollows:section2describesthedata.Section3and4showshow
World Cup soccer matches led to drops in trading activity. Section 5 presents evidence how,
simultaneously, price formation on national stock markets decoupled from global markets.
Section6presentsadditionalresultsandrobustnesschecks,whilesection7concludes.

7
2.Dataandmethodology
The dataset contains information on trading in major stocks on stock exchanges in nine
European countries, four countries from Latin America and one country each from North
AmericaandAfrica,onaminutebyminutebasis.Weusetworequirementswhenchoosingthe
countries.First,highfrequencystockmarketdatahavetobeavailable,and,second,thenational
teamhadtoplayatleastonesoccermatchduringtradinghoursofthenationalstockmarket.Per
country, we use the major stock market index to define which individual stocks we would
include.ThestockmarketdatawasobtainedthroughBloomberg.
Table1presentsanoverviewof thestockmarket indicescoveredforeach country,and
the number of stocks contained in each index. The number of stocks differs vastly across
countries, ranging from 17 in Argentinas MERVAL Index to 100 in the UKs FTSE100 Index. In
the light of this, and due to different market depths across countries, it will be important to
normaliseourtradingactivityindicatorsbycountry,aswewouldobviouslyexpectmoretrades
andhighervolumesinagivenminuteifweaggregateoverrelativelymorestocks,andfordeeper
markets. The final two columns list the trading hours of the national stock exchange, both in
localtimeandinSouthAfricantime.
Table1
For each stock contained in the most relevant stock market index, we collected
information on the number of trades, trading volumes (measured as the number of traded
stocks),thenumberofbidaskquotesandthequotedvolumesonaminutebyminutebasis.We
aggregatethisinformationtothenationallevel,suchthatwecovertheentiretradingactivityfor
the most relevant stocks in a given country. Furthermore, we also obtained the level of the
national stock market index on a minutebyminute basis. We are only looking at the largest
stocks in each country, as data availability poses less of an issue. This choice might introduce a
bias against finding drops in trading activity. Any reduction in trading activity might be
disproportionatelylargerforlocalisedstockswhichtradeinthinnermarkets.
Duringthe2010FIFAWorldCup,thenationalteamofoneofthecountriesinour
database was playing during regular national trading hours on 29 occassions. Table 2 has
informationontherelevantmatches.Capitallettersindicateweareabletoevaluateeffectsfora
particularmatch.Wecover21matches,asineightcases,dataforbothteamsareavailable.Table
2 shows how most matches in our sample are from the group stage. Also, we include two
secondroundmatchesandonequarterfinal.

Table2

8
In addition to analysing own matches matches where the national team of a given
countryisplayingwetestwhethertradingactivityisaffectedwhenanyoftheother31nations
areplaying.Inthiscase,thenumberofeventsincreasesto317.
9
Atthesametime,giventhatthe
nationalteamisnotparticipating,wewouldexpectsmaller,ifany,effectsontradingactivity.
ToestimatetheeffectofWorldCupmatchesontradingactivity(measuredbyeitherthe
numberoftradesorthevolumeoftradedstocks),wefirstdefineexpectedtradingactivity, A T ,
incountrycatthetimeofdaytandtheweekdaydowas:
, ,
, ,
, ,
,
c t dow
t
c t dow
c t dow
TA
TA t WorldCup
n
= e

, (1)
where TA denotes actual trading activity and n is the number of observations. Our
measurethereforecomputestheaveragetradingactivityobservedinagivencountryforagiven
timeofday,andseparatelyforeachdayoftheweekoutsidetheWorldCup( Cup World t e ).As
such, this benchmark controls for country effects, dayoftheweek effects and timeoftheday
effects. To compute this benchmark, we use six weeks prior to the World Cup and three weeks
afterwards. The 2010 World Cup started on June 11 and ended on July 11. Our full dataset
covers,therefore,theperiodfromMay1
st
toJuly31
st
,2010.
Then, we compute abnormal trading activity (ATA) as the percentage difference from
expectedactivity:
, , , ,
, ,
, ,
100
c t dow c t dow
c t dow
c t dow
ATA TA
ATA
TA

= , (2)
So,foreachminuteduringaWorldCupmatch,wecancompare,say,thenumberoftradestothe
averagenumberoftradesonthatparticularcountrysstockexchange,duringthesametimeof
day, during the same weekday, but during a period outside the World Cup. In the empirical
analysis, we use these measures of abnormal activity (either trades or volumes) as our
dependent variables. In the robustness analysis, we show that using an alternative measure of
abnormaltradingactivityleadstoqualitativelysimilarconclusions.

9
Notethathere,weareabletousematchesbyallcountriesparticipatingintheWorldCup,andnotjust
thefifteenforwhichwewereabletoobtainhighfrequencydata.Theonlyrequirementisthatthese
matchestakeplaceduringtradinghoursinthecountriesforwhichwehavedata.

9
3.Lesstradingactivityduringsoccermatches
Wefirstshowgraphicalevidence.Figures1and2presenthistogramsoftherelativenumberof
trades and trading volumes as defined in equation (2). These variables are shown for the
national matches (black bars) and for the benchmark period before and after the World Cup
(grey bars). The latter only includes those times of the day and days of the week where we
actually observe matches, meaning we use directly comparable data.
10
Figure 1 presents the
resultsforthepooleddataset,whereasFigure2showscountryresults.
Figures 1 and 2 suggest the following three points. First, when the national team is
playing, there is a shift in the distribution of abnormal trading activity. Both for the number of
trades and trading volumes, the figures indicate lower levels of market activity. Second, our
variablesofinterestarenonnormal,whichweneedtotakeintoaccountinourestimations.Our
mainstrategyistousemedianregressionsratherthanleastsquaresregressions,tocounterthe
effect of outliers. In one of the robustness exercises, we use tobit regressions, which take into
accountthatthedependentvariablesaretruncatedfrombelowat100%.Wefurthermoreshow
how the findings are robust to a number of other estimation methods. Third, the reduction
portrayed in Figure 1 is mirrored in the national breakdown, with evident reductions in nearly
allcountries.Atthesametime,thereisimportantheterogeneityacrosscountries.
Figures1and2
Table3providesastatisticalassessmentoftheshiftsintradingactivity.Weshowresults
for numbers of trades and trading volume, first for the pooled data, second broken down by
country,andfinallyaggregatedbycontinent.
11
Column1ofTable3providesthemediannumber
oftradesandthemediantradingvolumeduringourcontrolsample,sobeforeJune11andafter
July 11, 2010, during time windows that match the time windows of the soccer matches. The
second column shows the reduction in trading activity during other nations matches (so,
excluding matches in which the national team participated), and the third column contains the
corresponding numbers for own matches. Column 4 indicates whether the reduction in trading
activityduringownmatchesissignificantlylargerthanduringothernationsmatches.
Tables3and4
Startingwiththepooledresults,theimpressionfromfigures1and2isclearlyconfirmed.
Marketactivitydropssignificantlyduring2010WorldCupsoccermatches.Themediannumber

10
For these plots, but not for the later analysis, we exclude values larger than 500%. Such observations
arerareandmakethereadingofthechartdifficult.
11
Note,though,thatforNorthAmericaandAfrica,theseresultsarebasedonjustonecountry.

10
oftradesislowerby24%duringothernationsmatches,andby45%duringownmatches.Not
surprisingly, this effect is substantially larger if a countrys national soccer team is playing
compared to matches by other nations. Whereas a reduced number of transactions could be
compensated by an increase in the volume traded per transaction, this is clearly not the case:
alsothemedianvolumeisreduced,andevenbyasomewhatlargermagnitude(namelyby55%).
Again, there are substantial differences across countries. Still, in nearly all countries,
tradingactivityisstronglyreduced.Thedropintradingactivitiesforindividualcountriescanbe
substantial, with a maximum of 40% reduction in the median number of trades for other
nationsmatches(incaseofArgentina),andamaximumof83%fortheownmatches(incaseof
Chile). The Latin American countries show particularly large declines for the case of the
reduction in number of trades during own matches, the four Latin American countries are
amongthesixlargestreductionsobserved.Inlinewiththis,Table4testsfordifferencesacross
continents, and finds that, indeed, Latin America shows the biggest drop in trading activity
duringownmatches.
AfinalcommentrelatestotheresultsfortheUnitedStates.Contrarytoourprior,wealso
findstrongindicationsofdecliningactivityinU.S.stockmarkets.Often,theUnitedStatesisnot
perceived as being enthusiastic about soccer. At the same time, investors in the U. S. markets
oftenhaveaninternationalbackground,whichmightexplaintheratherstrongeffects.Thisrole
of international investors is also crucial in Kaplanski and Levys (2010) argument that the loss
effect after World Cup matches has had negative effects on U.S. markets. Also, it must be said
that interest in soccer in the U.S. has been increasing over the years. As estimated by Nielsen,
well over 100 million people watched at least part of the games of the 2010 edition.
12
In
addition,theresultsbyLozano(2011)arerelevant,whoshowedthatevenintheUnitedStates,
workerstaketimeofftofollowmatches.

4.Matcheventsinfluencedtradingactivity
The median regressionsindicateshiftsintrading activity acrosstheentiretimeofthe matches.
Foramore detailedanalysis,weareinterestedhowtradingactivityevolvedduring,before and
afterthesoccermatches.Inparticular,canweshowthatitreactedtomatchrelatedevents,such
as goals, cautions or halftime? To code when the match occurred, we use match reports

12
http://blog.nielsen.com/nielsenwire/consumer/increasedviewershiponlinevisitsandad
engagementamongworldcuphighlights/.

11
availableontheFIFAwebsite.
13
Thesereportsidentifytheexactlengthofeachhalf,byproviding
theamountofinjurytime,andshowthetimingofspecialeventssuchasgoalsorredandyellow
cards.
14
Aswecannotbefullycertainthatthematchstartedontime,andashalftimemayhave
taken longer than the standard fifteen minutes, we use dummies for goals and cautions that
equaloneduringtheminutewhentheeventisrecorded,butalsoduringthetwominutesbefore
and after. The dataset comprises 63 goals and 142 red and yellow cards. If, for a given match,
both teams are in our dataset, we count these events for both countries. As explanatory
variablesfortheevolutionoftradingactivity,wealsoincludedummiesforthe15minuteperiod
priortothekickoff.Duringthisperiod,gamepreviewsaretelevised,andthenationalanthems
are being played, which may already distract investors. We also include dummies for halftime,
toseeiftradersmakeupforlostactivityduringthefirsthalf.Finally,wetestwhetherandhow
market activityrecovers,byaddingdummyvariablesforthefirst,thesecondandthirdquarter
ofanhouraftereachmatch.Theseregressionsdothereforecontainobservationsforthematch
time as well as the preceding fifteen and the subsequent 45 minutes. The regression is (for the
exampleoftradesperminute)specifiedas:
15

, 1 2 , 3 , 4 , 5 ,
6 , 7 , 8 , ,
_1 _ 2 _ 3
c t c t c t c t c t
c t c t c t c t
trades goal card anthems halftime
aftermath aftermath aftermath u
o o o o o
o o o
= + + + + +
+ + +
(3)
Table 5 reports results. Columns 1 and 2 report benchmark results from median
regressions, whereas all subsequent ones report sensitivity tests, which we will discuss in
Section6.Theconstantterminthebenchmarkresultsindicatesthat,onaverage,mediantrading
activity is lower by 47% for the number of trades, and by 57% for the trading volume. These
magnitudesareinlinewithtable3.However,duringthefiveminutessurroundingagoal,there
is an additional reduction of around 5% (significantly estimated for the number of trades). In
contrast, cautions do not lead to additional drops in activity. Interestingly, the reduced market
activity starts already in the runup to the match. Again, the effects are sizeable, in the
magnitudeof40%fornumberoftrades,and50%fortradingvolume.Thesefiguresareobtained
asthesumoftheconstantandthecoefficientontheanthemsvariable.Afterthematch,trading
recovers, albeit slowly. Only after 3045 minutes, trading activity is roughly back to normal
levels.Duringthehalftimebreak,tradingactivityisstillaround35%lowerthanonbenchmark
days. However, market activity is higher than during actual playingtime, suggesting investors
docompensatefortheloweractivityduringthefirsthalf,andtheexpectedlossofactivityduring
thesecondhalf.

13
Source: http://www.fifa.com/worldcup/archive/southafrica2010/matches/index.html. See also:
http://www.fifa.com/mm/document/affederation/technicaldevp/01/29/30/95/reportwm2010_web.pdf
14
Asthereisonlyonepenaltyinoursample,wecannotcheckforanyeffects.
15
Thedependentvariableisasdefinedinequation(2),sowealreadycontrolforcountryfixedeffects.

12
Table5

5.Decoupling:lesscomovementwithglobalstockmarkets
Having established that there is reduced trading activity during soccer matches, we will now
investigatewhetherthisalsohadeffectsonpriceformationconsistentwithareducedattention
due to the distraction presented by the matches.
16
The main results are based on an aggregate
approach, by studying stock indices rather than prices of individual stocks. However, section 6
will report additional findings on the dispersions of returns of individual stocks. We are
specifically interested in the effects of matches played by the national team on the countrys
stock exchange. If investors are less attentive during matches, relevant news would not be
incorporated as quickly in prices as under normal circumstances (in line with previous results
by DellaVigna and Pollet (2009), Hirshleifer, Lim and Teoh (2009) or Louis and Sun (2010)).
Unfortunately, there is no systematic news arrival during our sample (such as earnings
announcementsorthereleaseofmacroeconomicdata)thatwouldallowforadirecttestofthis
hypothesis.
Accordingly,wewilltakeanindirectapproacheandstudythecomovementbetweenthe
national stock market and the global stock market. Under the assumption that the global stock
market should be affected only marginally by a Wrold Cup soccer match, we would expect that
the global stock market continues to price news in a regular fashion. If attention really had
sizeableeffectsonthenationalstockmarkets,wewouldexpectthatthiscandaffectthenational
pricingpatterns,leadingtoaneglectofnewsthatgetpricedintotheglobalmarket.Accordingly,
the comovement between national and global markets would be less strong.
17
To test this
hypothesis,weestimatethefollowingmodelusingminutebyminutedata:
, 1 2 , 1 3 , 2 4 , 3 , , ,
1 , 2 , 3 , ,
c t c t c t c t w t w t c t
c t c t c t c t
r r r r r r match
match dow tod u
o o o o | o


= + + + + + +
+ + +
(4)

16
WealsoassessedwhetherthelosseffectdocumentedbyEdmansetal.(2007)couldbeconfirmedusing
our highfrequency data. To this end, we computed abnormal returns and regressed those on dummies
measuringwinsandlosses.Forhorizonsupto24hoursafterthematch,wefoundnosignificanteffectsof
matchoutcomesonreturns.However,itshouldbenotedthatthenumberofmatchesissmallcomparedto
Edmans et al.. Also, most of the matches included in our sample were played during the group stage,
wheretheimpactoflosingamatchissmallerthanduringtheeliminationstage.
17
Manyauthorshaveanalyzedthecomovementofstockmarketsacrosscountries.Oneissueinthedebate
iswhetherthecomovementofreturnsacrossnationalmarketshasincreased,andifso,why.Another
pointofdiscussioniswhethercountryorindustryfactorsarethemostrelevantdriverofcomovement
(Bekaert,HodrickandZhang2009;BrooksandDelNegro2004;HestonandRouwenhorst1994).

13
where r
c,t
denotes index returns in country c, r
w,t
is the return of the global stock market index
(measuredbytheMSCIWorldIndex),match
c,t
isadummyvariableequaltooneduringthetime
of a national teams match, dow
c,t
are dummy variables for weekdays (Monday through
Thursday),andtod
c,t
aredummyvariablesforeachfiveminuteperiodofaday.
18
Intheextreme
case, if only local investors trade on the stock exchange, and assuming full inattention, there
wouldbenopricemovementsinlocalmarkets.Inthatcase,ourestimationswouldindicatethat
=.Inpractice,aportionofthetradesisinitiatedbyinternationalinvestors.Also,presumably,
onlyafractionofinvestorswillbedistractedbysoccermatches.Therefore,weexpect0<<.
We estimate the regressions using panelcorrected standard errors (PCSE) proposed by
BeckandKatz(1995).Incontrasttorelatedwork(Edmansetal.2007,HirshleiferandShumway
2003), we assume that the only deviations from Gaussian errors in u
c,t
are due to panel
heteroskedasticity. This means that we allow the variance of u
c,t
to be countryspecific, but do
notallowforcontemporaneouscrosscountrycorrelation.Thereasonisthatthematchesinour
sampleareusuallynotplayedatthesametime.
19

Forbrevity, Table6showsonlyselectedcoefficients.Theparametersofinterestarethe
comovement of stock returns when the national team is not playing a match (coefficient |,
labelled as beta in the table) and the changing strength of this comovement during matches
(coefficiento,labelledasdecouplinginthetable).Forthepooledregressionreportedinthefirst
row, beta is estimated to be 0.87, indicating a high degree of international stock market
integration.
20
Interestingly, however, this comovement became substantially weaker during
national team matches: On average, the comovement is reduced by 21%. On the one hand, this
suggeststhatnationalstockmarketsstillfollowglobaldevelopmentsduringmatches.However,
thereisastrongsenseofdecoupling,asmarketsarefollowingdevelopmentsonthesoccerpitch,
ratherthaninthetradingpit.
ThecountrybycountryregressionsintheremainingpartofTable7confirmthisoverall
picture.Evenifthedecouplingparameterissignificantlyestimatedonlyforsevensixcountries,

18
Allowingforlagsoftheglobalstockmarketindexorestimatingastaticmodelleadstosimilar
conclusions.Resultsavailableonrequestfromthecorrespondingauthor.
19
An alternative estimation method would be the FGLS approach (Parks 1967, Kmenta 1986). Beck and
Katz(1995)arguedthatFGLSmayproduceanticonservativestandarderrorsincaseswherethenumber
oftimeobservations(T)issmallrelativetothenumberofpanels(N).Foreachofthefifteencountriesin
oursample,however,atleast85minutesofplayingtimeduringmarketopeninghoursareavailable.Still,
Beck and Katz (1996) have also argued that the efficiency gains of FGLS will be minor unless high
heteroskedasticityandparameterhomogeneityarepresent.
20
By construction, the MSCI World Index is to a large extent driven by developments in U.S. and U.K.
markets.Therefore,wereranthepooledregressionwithoutthesetwocountries,whichgaveverysimilar
results.

14
thepointestimatesarenegativeformostcountries.Germanyisthenotableexception,wherewe
find,counterintuitively,anincreaseinthestrengthofthecomovement.
Table6
6.Robustnessandextensions
This section reports on a large number of robustness tests and various extensions. We will
proceed along the three key findings. First, we discuss the results of reduced trading activity
during soccer matches, then proceed to the reactions to match events, and finally discuss the
resultsforpriceformation.

6.1Reducedtradingactivityduringsoccermatches:robustness
Usingmarketmakerstomeasureactivity,andusingdifferentestimators
Table 7 reports on a number of robustness checks for the results of Table 3. Whereas we had
previously tested for effects on the number of trades and the traded volumes, the row Quotes
repeats the estimations for the number of quotes given per minute, as well as the quoted
volumes.Thefocushereisonmarketmakeractivity,ratherthanonactualtrades.Again,wefind
strongindicationsofreducedactivityduringnationalteammatches.
Second, given the nonnormality of the variables of interest, we had used median
regressionsintable3.TherowOLSregressionintable7showshowthequalitativeconclusions
aresimilarwhenusingleastsquaresregressions(usingpanelcorrectedstandarderrors).Third,
as our dependent variable is truncated from below at 100%, we also ran tobit regressions.
Again,theconclusionisthatthereisasignificant reduction in thenumber oftradesandtraded
volumes.
Table7
EvidenceusingGermangovernmentbondfutures
A further robustness test checks whether the results obtained are specific to the 2010 FIFA
WorldCup,orcouldalsobeobservedmoresystematically.Unfortunately,wedonothaveaccess
to the equivalent intraday stock market data that would allow expanding our analysis to
previous World Cups. Instead we reverted to an intraday dataset for trading in longterm
German government bond futures, covering a highly liquid basket of both nonbenchmark and
benchmark German governments bonds with a remaining timetomaturity between 8.5 and

15
10.5 years. We have obtained these data from TickData.
21
Regarding the number of trades, our
data covers the World Cups organized by France (1998), Japan and South Korea (2002),
Germany (2006) and South Africa (2010). We evaluate shifts in investor attention for a total of
eightmatchesduringthesetournaments,inwhichGermanyparticipated,andwhichwereplayed
duringGermantradinghours.Regardingtradingvolumes(measuredasthenumberofcontracts
over 1 million each) our data sample is shorter, such that we are able to analyze effects for
three German matches during the World Cups in 2006 and 2010. This robustness test implies
thatwecoveraverydifferentfinancialmarket,aswellassoccermatchesplayedduringdifferent
WorldCups.Table7showthattheeffectsareagainsizeable,withareductionintradingactivity
of29%forthenumberoftrades,and48%forthetradingvolume.
22

Placebotimeschedule
In order to assess whether the results are truly driven by the soccer matches, or possibly by
other factors, we ran the empirical analysis, while shifting the times of the soccer matches by
120 minutes backward. So, a match that started at 16:00 would be coded to start at 14:00. Of
course,wewouldexpecttoseenoeffectoftheseartificialmatchesontradingactivity.Indeed,as
thefinalrowoftable7shows,theeffectissmallandstatisticallyinsignificantforthenumberof
trades, and virtually zero for the volume traded. We also applied the placebo time schedule to
our analysis of decoupling. Doing so, we find as expected that the decoupling parameter is
insignificant.
23

Alternativedefinitionofabnormaltradingactivity
In our benchmark analysis, we compare trading activity during matches to average activity
during corresponding days of the week and times of the day outside the World Cup. In this
section, we study abnormal trading activity using a filtering procedure as in Edmans et al.
(2007),whofollowGallant,RossiandTauchen(1992).Theideaistoconstruct,foreachcountry,
ameanzero,unitvarianceseriesforabnormaltradingactivity
ct
.First,foreachcountryinthe

21
The data generally refer to the contract with the nearest maturity. The switch to the next maturity is
done by a procedure that compares daily tick volumes for two adjacent contracts. It switches usually
around 35 days before expiration of the contract with the nearest maturity, when daily tick volumes
exceed those of the old contract. This procedure ensures maximum liquidity of the considered contracts.
Formoreinformation,seehttp://www.tickdata.com.
22
If we exclude the one German match from the 2010 World Cup, the number of trades still shows a
significant decline of 22.4%. On the other hand, there is no significant decline in the traded volumes,
possiblyduetothefactthatthesamplecoversonlytwomatchesinthiscase.
23
Resultsnotshown;availableuponrequestfromcorrespondingauthor.

16
dataset, we run the regression V
ct
=
0c
x
ct
+ u
ct
, where V
ct
denotes the natural logarithm of
tradingactivity(eitherthenumberoftradesortradingvolume)incountrycinminutet,andx
ct
containsaconstant,timeofdaydummiesforeachfiveminuteperiodduringtradinghours,day
oftheweekdummies,weeklyandmonthlydummies,atimetrend,thetimetrendsquared,and
15lagsofthedependentvariables.
24
Next,weusetheresidualstomodelthevarianceaslog(
2
ct
)
=
1c
y
ct
+
ct
where y
ct
has the same components as x
ct,
with the exception of the lagged
dependentterms.Intheend,abnormaltradingactivityisdefinedas
ct
=a
c
+b
c

ct
/exp(
1c
y
ct
/2)

wherewechoosea
c
andb
c
suchthat
ct
haszeromeanandunitvariance.
As these alternative measures of abnormal trading activity show few outliers, we use
leastsquaresregressions,againusingpanelcorrectedstandarderrors.Thebottomrowoftable
7hasresults,whichbroadlyconfirm ourearlier findings.During matches bythe nationalteam,
there is a reduction in both measures of trading activity (column 3). For number of trades, the
abnormalreductionis0.14standarddeviations,whilefortradingvolumes,thereductionis0.16.
Inaddition,wealsofindasignificant,thoughsmall,reductionintradesduringmatchesinwhich
thenationalteamdidnotparticipate.

6.2Robustnessanalysisformatchevents
Asmentionedinsection4,weperformedvariousrobustnessresultsfortheanalysisofmatch
events.Inadditiontothebenchmarkfindings,theremainingcolumnsoftable5listresults
replacingtradesbyquotes(panelQuotes),conductingleastsquaresregressions(panelOLS
regression),tobitregressions(panelTobitregression),artificiallyshiftingbackwardthe
timingofallmatchesby120minutes(panelShiftedtimeschedule),andusingthealternative
measureofabnormaltradingactivity(panelAlternativedefinitionofabnormaltrading).Inthe
firstthreecases,theconstanttermscontinuetopointtoamarkeddropintradingactivityduring
matches.Also,theadditionaleffectofgoalsscoredbyeitherteamonthenumberoftradesis
replicatedbytheotherestimations.Whenusingtheplacebotimeschedule,wefindnoclear
significantfindings.Inthelastcolumn,usingthealternativemeasureofabnormaltrades,wefind
thatgoalsareaccompaniedbysignificantdeclinesinthenumberoftrades(of0.19standard
deviations),butdonotsignificantlyaffecttradingvolumes.
25

24
Wealsoused,withsimilarresults,specificationswithmoreandfewerlags.
25
Note that in this case, trading activity is already above match levels in the first 15 minutes after the
match ended. This is due to the construction of the abnormal trade levels, which use fifteen lags of the
variables. By treating the depressed levels of activity during the match as normal market circumstances,
evenaslightuptickintradingafterthematchendswillbeseenasanabnormalincrease.

17
6.3Furtherresultsforpriceformation
Decoupling:salienceofinformation
The literature has identified that the salience of information may matter in the presence of
inattentive investors, with more salient information receiving relatively more attention (e.g.
Barber and Odean 2008). Our setup allows for testing this by conditioning on the size of price
changes. Thus, we study whether the reduction in comovement is more pronounced for large
thanforregularsizedpricemovementsinglobalmarkets.Largepricechangesinglobalmarkets
(ortheunderlyingnewsthattriggeredthesepricechanges)aremorelikelytogetnoticed,even
by relatively inattentive investors. In contrast, regular swings in global prices might be more
easily overlooked, especially when soccer matches are also drawing attention. Accordingly, we
expect that the decoupling results presented above are particularly pronounced in a situation
whenpricechangesinglobalstockmarketswererelativelymuted.
Totestthis hypothesis, weextendtheregressionmodelofequation4bydifferentiating
both the beta and decoupling terms depending on the magnitude of the movement in global
stock markets. Table 8 presents the corresponding results. In the first column, we classify
returns in the global stock market as large, if they are above the 90
th
and below the 10
th

percentile of the sample distribution. In the second column, the definition is based on
movementsbeyondthe85
th
andthe15
th
percentile.
Table8
Two findings emerge. When there are no soccer matches in progress, the comovement
between national and global stock markets differs somewhat across large and regularsized
market movements. In particular, the comovement is smaller when we condition on large
movements in global stock markets. Second, in line with our hypothesis, the decoupling is
particularly strong during regular price changes on global markets. In the presence of salient
price swings in global stock markets, we still find evidence for a lower comovement between
globalandlocalmarkets.However,thedegreeofdecouplingislesspronounced.Thecoefficient
isonlyreducedbysome15%to18%.Whenswingsinglobalstockpricesarerelativelysmaller
(and thus less salient), the degree of decoupling is estimated to be well over 40%. This low
degree of comovement suggests that national stock markets pretty much dance to their own
tune during soccer matches, at least as long as there are no large movements in global stock
markets.Thedifferentdecouplingresultsareinlinewiththeideathatsalientinformationhasa
greater impact on price formation, thus confirming the hypothesis that there is investor
inattentionduringsoccermatches.

18
Decouplingduringlunchhours
Does the drop in trading activity really constitute a shift in attention? To provide further
guidanceonourresults,westudywhathappensduringlunchhours,asthisisanotheroccasion
on which trading activity is often systematically lower. At the same time, we would expect that
market participants arrange their lunch time in a way that ensures attention to the ongoing
developments,suchthatwewouldnotexpectadecouplingtooccur.Therefore,examiningtrade
andreturnpatternsduringlunchhoursservesasabenchmarktoseeifsoccerhasreallyactedas
adistraction,byloweringattention.
Wedefinelunchtimeasthetwohourswiththelargestdropintradingactivityrelativeto
therest ofthetradingday.Forinstance,for mostofthe Europeancountriesinoursample,this
definition takes the two hours between 12:00 and 13:59 as the lunch break. Using the same
approachasforthesoccermatches,wefindthatthemediantradingactivityisreducedby38%
(43%) for the number of trades (volume traded). This decline in trading activity is comparable
towhatweestablishedforsoccermatches,andsupportsusinglunchhoursasabenchmark.
How are stock returns affected during lunch breaks? To examine this, we extend the
regressionmodelofequation(4)to
, 1 2 , 1 3 , 2 4 , 3 , 1 , , 1 , ,
1 , 2 , 3 , 4 , ,
c t c t c t c t w t w t c t w t c t
c t c t c t c t c t
r r r r r r match r lunch
match dow tod lunch u
o o o o | o o


= + + + + + + +
+ + + +
(5)
whereallvariablesaredefinedasinequation(4),withtheadditionofadummyvariable
that is equal to one during lunchtime, and its interaction with the global stock market index.
Estimation results are given in Table 9. First, including the dummy for lunch hours does not
changetheestimatesfortheothervariables.Second,wefindasmalleffectoflunchhoursonthe
comovement between local and global stock returns. However, compared to the decoupling
during soccer matches, the decoupling parameter of 0.018 during lunch hours is negligible. In
comparison, the large decoupling during soccer matches indicates that inattention had
substantivepriceeffects.
Table9

Areducedpricingoffirmspecificinformation
A final test exploits the crosssectional variation in our dataset. Peng and Xiong (2006) have
shown that limited investor attention leads investors to process more market and sectorwide

19
information than firmspecific information. Applying this finding to our case, it should be that
individualstockpricesmoveinalessidiosyncraticfashionduringsoccermatches,i.e.thatthere
is stronger comovement across the stocks in a given country. To test this hypothesis, we
computed the standard deviation of minutebyminute stock returns across all constituents of
the national stock index. Table 10 shows that percentage difference of this dispersion measure
during soccer matches compared to the benchmark period outside the World Cup (defined
analogously to equations 1 and 2). We find a significant reduction in dispersion during other
nations matches of 19.5%, and even a slightler stronger drop during national team matches.
Thisresultisinlinewiththehypothesisthatduringtimesofinattention,firmspecificnewsare
pricedintoasmallerextentthanmarketandsectorwideinformation.
Table10
7.Conclusions
This paper uses a major sporting event to study the occurrence and effects of shifts in investor
attention. Using highfrequency data on fifteen international stock markets, we present three
pieces of evidence on reduced attention during soccer matches at the 2010 FIFA World Cup.
Usingvariousmeasures,primarlythenumberoftransactionsandthevolumesoftradedstocks,
we find evidence of strongly reduced activity in stock markets, especially during matches in
which the national team competed. Using minutebyminute data also allows us to relate stock
market activity to events during matches. In particular, we find how the number of offered
quotes and the number of actual trades were significantly lower at the time when goals were
scored.
Inlinewithpreviouswork(DellaVignaandPollet2009;Hirshleifer,LimandTeoh2009;
Louis and Sun 2010), we show how limited attention in stock markets has had implications for
priceformation.Nationalstockmarketscomovedlesswithglobalstockmarkets,suggestingthat
newswhichgotpricedintotheglobalmarketaffectedthenationalmarketinadifferentfashion
than otherwise. This decoupling was especially large when there were no outstanding price
movements in global stock markets, which indicates that in particular less salient information
receivedlessattentionduringsoccermatches.Interestingly,nosuchdecouplingisfoundduring
lunchtime, despite a reduction in trading activity of a similar magnitude. In the light of this, we
conclude that markets were following developments on the soccer pitch rather than in the
tradingpit,leadingtoachangedpriceformationprocess.

20
References
Baade,R.A.andV.A.Matheson(2004)TheQuestfortheCup:AssessingtheEconomicImpactof
theWorldCupRegionalStudies38(4),343354.
Barber,B.M.andT.Odean(2008)AllThatGlitters:TheEffectofAttentionandNewsonthe
BuyingBehaviorofIndividualandInstitutionalInvestorsReviewofFinancialStudies
21(2),785818.
Beck,N.andJ.N.Katz(1995)Whattodo(andnottodo)withtimeseriescrosssectiondata
AmericanPoliticalScienceReview89(3),634647.
Beck,N.andJ.N.Katz(1996)Nuisancevs.Substance:SpecifyingandEstimatingTimeSeries
CrossSectionModelsPoliticalAnalysis6(1),136.
Bekaert,G.,R.J.HodrickandX.Zhang(2009)InternationalStockReturnComovementJournal
ofFinanceLXIV(6),25912626.
Brooks,R.andM.DelNegro(2004)Theriseincomovementacrossnationalstockmarkets:
MarketintegrationorITbubble?JournalofEmpiricalFinance11,649680.
Cohen,L.andA.Frazzini(2008)EconomicLinksandPredictableReturnsJournalofFinance
LXIII(4),19772011.
Corwin,S.A.andJ.F.Coughenour(2008)LimitedAttentionandtheAllocationofEffortin
SecuritiesTradingJournalofFinanceLXIII(6),30313067.
DellaVigna,S.andJ.M.Pollet(2007)DemographicsandIndustryReturnsAmericanEconomic
Review97(5),16671702.
DellaVigna,S.andJ.M.Pollet(2009)InvestorInattentionandFridayEarningsAnnouncements
JournalofFinanceLXIV,709749.
Edmans,A.,D.Garcaand.Norli(2007)SportsSentimentandStockReturnsJournalof
FinanceLXII(4),19671997.
Gabaix,X.,D.Laibson,G.MolocheandS.Weinberg(2006)CostlyInformationAcquisition:
ExperimentalAnalysisofaBoundedlyRationalModelAmericanEconomicReview96(4),
10431068.
Gallant,R.,P.E.RossiandG.Tauchen(1992)StockpricesandvolumeReviewofFinancial
Studies5,199242.

21
Hagn,F.andW.Maennig(2008)EmploymenteffectsoftheFootballWorldCup1974in
GermanyLabourEconomics15,10621075.
Heston,S.L.andG.K.Rouwenhorst(1994)Doesindustrialstructureexplainthebenefitsof
internationaldiversification?JournalofFinancialEconomics36,327.
Hirshleifer,D.,andT.Shumway(2003)GoodDaySunshine:StockReturnsandtheWeather
JournalofFinanceLVIII(5),10091032.
Hirshleifer,D.andS.H.Teoh(2003)Limitedattention,informationdisclosure,andfinancial
reportingJournalofAccounting&Economics36,337386.
Hirshleifer,D.,S.S.LimandS.H.Teoh(2009)DriventoDistraction:ExtraneousEventsand
UnderreactiontoEarningsNewsJournalofFinanceLXIV(5),22892325.
Hou,K.L.PengandW.Xiong(2009)Ataleoftwoanomalies:Theimplicationsofinvestor
attentionforpriceandearningsmomentumWorkingPaper.Availableat
http://ssrn.com/abstract=976394.
Kahnemann,D.(1973)AttentionandEffort,EnglewoodCliffs,NJ:PrenticeHall.
Kaplanski,G.andH.Levy(2010)ExploitablePredictableIrrationality:TheFIFAWorldCup
EffectontheU.S.StockMarketJournalofFinancialandQuantitativeAnalysis45(2),535
553.
Kmenta,J.(1986)ElementsofEconometrics2
nd
edition.NewYork:MacMillan.
Louis,H.andA.Sun(2010)InvestorInattentionandtheMarketReactiontoMerger
AnnouncementsManagementScience56(10),17811793.
Lozano,F.A.(2011)TheflexibilityoftheworkweekintheUnitedStates:Evidencefromthe
FIFAWorldCupEconomicInquiry49(2),512529.
Mackowiak,B.andM.Wiederholt(2009)OptimalStickyPricesunderRationalInattention
AmericanEconomicReview99,769803.
Mankiw,G.andR.Reis(2002)StickyInformationVersusStickyPrices:aProposaltoReplacethe
NewKeynesianPhillipsCurveQuarterlyJournalofEconomics117,12951328.
Mondria,J.,T.WuandY.Zhang(2010)Thedeterminantsofinternationalinvestmentand
attentionallocation:UsinginternetsearchquerydataJournalofInternationalEconomics
82,8595.

22
Parks,R.(1967)Efficientestimationofasystemofregressionequationswhendisturbancesare
bothseriallyandcontemporaneouslycorrelatedJournaloftheAmericanStatistical
Association62(318),500509.
Peng,L.andW.Xiong(2006)Investorattention,overconfidence,andcategorylearningJournal
ofFinancialEconomics80(3),563602.
Rose,A.K.andM.M.Spiegel(2011)TheOlympiceffectEconomicJournal,121(553),652677.
Sims,ChristopherA.ImplicationsofRationalInattentionJournalofMonetaryEconomics,50(3),
66590.
SkogmanThoursie,P.S.(2004)Reportingsick:aresportingeventscontagiousJournalof
AppliedEconometrics19,809823.
White,H.(1980)AHeteroskedasticityConsistentCovarianceMatrixandaDirectTestfor
HeteroskedasticityEconometrica48,817838

23
Figure1a:Abnormalnumberoftrades,forcontrolsampleandduringownmatchesatthe2010
WorldCup
0
5
1
0
1
5
-100 0 100 200 300 400 500

Figure1b:Abnormaltradingvolume,forcontrolsampleandduringownmatchesatthe2010
WorldCup
0
5
1
0
1
5
-100 0 100 200 300 400 500

Note: Percentage differences of trading activity compared to the average activity in the same country, at
thesametimeofdayandthesamedayoftheweekforacontrolsamplecoveringseveralweeksbeforeand
afterthe2010WorldCup.Seeequations1and2inthemaintext.Greybarsrepresentdataforcomparison
timewindows,blackbarsfortimewindowsduringWorldCupmatchesbythenationalteam.

2
4

F
i
g
u
r
e

2
a
:

H
i
s
t
o
g
r
a
m

o
f

a
b
n
o
r
m
a
l

t
r
a
d
e
s
,

f
o
r

c
o
n
t
r
o
l

s
a
m
p
l
e

a
n
d

d
u
r
i
n
g

o
w
n

m
a
t
c
h
e
s

a
t

t
h
e

2
0
1
0

W
o
r
l
d

C
u
p
,

b
y

c
o
u
n
t
r
y

0 1 0 2 0 3 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
A
r
g
e
n
t
i
n
a
0 1 0 2 0 3 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
B
r
a
z
i
l
0 1 0 2 0 3 0 4 0 5 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
C
h
i
l
e
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
D
e
n
m
a
r
k
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
E
n
g
l
a
n
d
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
F
r
a
n
c
e
0 1 0 2 0 3 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
G
e
r
m
a
n
y
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
I
t
a
l
y
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
M
e
x
i
c
o
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
N
e
t
h
e
r
l
a
n
d
s
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
P
o
r
t
u
g
a
l
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
o
u
t
h
A
f
r
i
c
a
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
p
a
i
n
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
w
i
t
z
e
r
l
a
n
d
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
U
n
i
t
e
d
S
t
a
t
e
s

N
o
t
e
:

T
h
e

c
h
a
r
t
s

s
h
o
w

t
h
e

p
e
r
c
e
n
t
a
g
e

d
i
f
f
e
r
e
n
c
e

o
f

t
r
a
d
i
n
g

a
c
t
i
v
i
t
y

c
o
m
p
a
r
e
d

t
o

t
h
e

a
v
e
r
a
g
e

a
c
t
i
v
i
t
y

i
n

t
h
e

s
a
m
e

c
o
u
n
t
r
y
,

a
t

t
h
e

s
a
m
e

t
i
m
e

o
f

d
a
y

a
n
d

t
h
e

s
a
m
e

d
a
y

o
f

t
h
e

w
e
e
k

f
o
r

c
o
n
t
r
o
l

s
a
m
p
l
e

c
o
v
e
r
i
n
g

s
e
v
e
r
a
l

w
e
e
k
s

b
e
f
o
r
e

a
n
d

a
f
t
e
r

t
h
e

2
0
1
0

W
o
r
l
d

C
u
p
.

G
r
e
y

b
a
r
s

r
e
p
r
e
s
e
n
t

d
a
t
a

f
o
r

c
o
m
p
a
r
i
s
o
n

t
i
m
e

w
i
n
d
o
w
s
,

b
l
a
c
k

b
a
r
s

f
o
r

t
i
m
e

w
i
n
d
o
w
s

d
u
r
i
n
g

W
o
r
l
d

C
u
p

m
a
t
c
h
e
s

b
y

t
h
e

n
a
t
i
o
n
a
l

t
e
a
m
.

2
5

F
i
g
u
r
e

2
b
:

H
i
s
t
o
g
r
a
m

o
f

a
b
n
o
r
m
a
l

t
r
a
d
i
n
g

v
o
l
u
m
e
s
,

f
o
r

c
o
n
t
r
o
l

s
a
m
p
l
e

a
n
d

d
u
r
i
n
g

o
w
n

m
a
t
c
h
e
s

a
t

t
h
e

2
0
1
0

W
o
r
l
d

C
u
p
,

b
y

c
o
u
n
t
r
y

0 1 0 2 0 3 0 4 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
A
r
g
e
n
t
i
n
a
0 1 0 2 0 3 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
B
r
a
z
i
l
0 2 0 4 0 6 0 8 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
C
h
i
l
e
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
D
e
n
m
a
r
k
0 5 1 0 1 5 2 0 2 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
E
n
g
l
a
n
d
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
F
r
a
n
c
e
0 5 1 0 1 5 2 0 2 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
G
e
r
m
a
n
y
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
I
t
a
l
y
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
M
e
x
i
c
o
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
N
e
t
h
e
r
l
a
n
d
s
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
P
o
r
t
u
g
a
l
0 5 1 0 1 5
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
o
u
t
h
A
f
r
i
c
a
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
p
a
i
n
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
S
w
i
t
z
e
r
l
a
n
d
0 5 1 0 1 5 2 0
-
1
0
0
0
1
0
0
2
0
0
3
0
0
4
0
0
5
0
0
U
n
i
t
e
d
S
t
a
t
e
s

N
o
t
e
:

T
h
e

c
h
a
r
t
s

s
h
o
w

t
h
e

p
e
r
c
e
n
t
a
g
e

d
i
f
f
e
r
e
n
c
e

o
f

t
r
a
d
i
n
g

a
c
t
i
v
i
t
y

c
o
m
p
a
r
e
d

t
o

t
h
e

a
v
e
r
a
g
e

a
c
t
i
v
i
t
y

i
n

t
h
e

s
a
m
e

c
o
u
n
t
r
y
,

a
t

t
h
e

s
a
m
e

t
i
m
e

o
f

d
a
y

a
n
d

t
h
e

s
a
m
e

d
a
y

o
f

t
h
e

w
e
e
k

f
o
r

c
o
n
t
r
o
l

s
a
m
p
l
e

c
o
v
e
r
i
n
g

s
e
v
e
r
a
l

w
e
e
k
s

b
e
f
o
r
e

a
n
d

a
f
t
e
r

t
h
e

2
0
1
0

W
o
r
l
d

C
u
p
.

G
r
e
y

b
a
r
s

r
e
p
r
e
s
e
n
t

d
a
t
a

f
o
r

c
o
m
p
a
r
i
s
o
n

t
i
m
e

w
i
n
d
o
w
s
,

b
l
a
c
k

b
a
r
s

f
o
r

t
i
m
e

w
i
n
d
o
w
s

d
u
r
i
n
g

W
o
r
l
d

C
u
p

m
a
t
c
h
e
s

b
y

t
h
e

n
a
t
i
o
n
a
l

t
e
a
m
.

26
Table1:Countryandstockmarketindexcoverage
Country StockIndex Numberof
stocksinindex
Tradinghours(local) Tradinghours
(SouthAfricantime)
Argentina MERVAL 17 11.0017.00 16.0022.00
Brazil BOVESPA 57 10.0017.00 15.0022.00
Chile IPSA 40 9.3016.30 15.3022.30
Denmark OMXCopenhagen20 20 9.0017.00 9.0017.00
England FTSE100 100 8.0016.30 9.0017.30
France CAC40 40 9.0017.30 9.0017.30
Germany DAX 30 9.0017.30 9.0017.30
Italy DowJonesItalyTitans30 30 9.0017.25 9.0017.25
Mexico IPC 20 8.3015.00 15.3022.00
Netherlands AEX 25 9.0017.30 9.0017.30
Portugal PSI20 20 9.0017.30 9.0017.30
SouthAfrica FTSE/JSE40 40 9.0017.00 9.0017.00
Spain IBEX35 35 9.0017.30 9.0017.30
Switzerland SMI 20 9.0017.30 9.0017.30
UnitedStates DJIA 30 9.3016.00 15.3022.00

Note:Thetableshowsthecountriescoveredinthedataset,thenameoftherespectivestockindices,and
thenumberofstockscontainedtherein.Thefinaltwocolumnsshowtradinghoursatthestockexchange
inlocaltimeandinSouthAfricantime.

27
Table2:Matchesduring2010WorldCupcoveredintheanalysis
Date Start Teams Score Redoryellow
cards
11Jun 16.00 SOUTHAFRICAMEXICO 11 4
14Jun 13.30 NETHERLANDSDENMARK 20 3
15Jun 16.00 CotedIvoirePORTUGAL 00 3
15Jun 20.30 BRAZILKoreaDPR 21 1
16Jun 16.00 SPAINSWITZERLAND 01 4
17Jun 20.30 FranceMEXICO 02 6
18Jun 13.30 GERMANYSerbia 01 9
18Jun 16.00 SloveniaUNITEDSTATES 22 5
21Jun 16.00 CHILESWITZERLAND 10 10
21Jun 13.30 PORTUGALKoreaDPR 70 4
22Jun 16.00 MEXICOUruguay 01 3
22Jun 16.00 FRANCESOUTHAFRICA 12 2
22Jun 20.30 GreeceARGENTINA 02 2
23Jun 16.00 UNITEDSTATESAlgeria 10 6
23Jun 16.00 SloveniaENGLAND 01 4
24Jun 16.00 SlovakiaITALY 32 8
25Jun 16.00 PORTUGALBRAZIL 00 7
25Jun 20.30 CHILESpain 12 4
28Jun 16.00 NETHERLANDSSlovakia 21 5
28Jun 20.30 BRAZILCHILE 30 5
02Jul 16.00 NETHERLANDSBRAZIL 21 6
Roundof16
Quarterfinals
Groupstage(round1)
Groupstage(round2)
Groupstage(round3)

Note: Start of the match according to time in South Africa. For the countries in capital letters, we can
evaluateeffectsonstockmarketsduringthatparticularmatch.

28
Table3a:Tradesperminute,andpercentagechangeduringWorldCupmatches
(1)Tradespermin,
outsideWorldCup
Comparison
(2)vs.(3)
All 142 23.792*** 0.281 45.152*** 0.955 +++
Argentina 4 40.000*** 1.432 72.093*** 4.697 +++
Brazil 725 17.061*** 0.708 65.312*** 0.821 +++
Chile 9 28.409*** 1.066 83.333*** 2.608 +++
Denmark 41 22.313*** 1.539 8.094 9.246 +++
England 1,370 23.108*** 0.927 21.274*** 1.917
France 1,312 20.523*** 0.965 30.266*** 3.383 +++
Germany 256 30.140*** 0.839 59.008*** 2.094 +++
Italy 637 21.857*** 0.929 20.473*** 2.477
Mexico 82 17.566*** 1.238 52.019*** 1.823 +++
Netherlands 347 24.213*** 1.211 29.236*** 2.694 +
Portugal 57 27.206*** 1.259 39.171*** 2.238 +++
SouthAfrica 13 5.954*** 1.990 54.123*** 4.408 +++
Spain 424 22.738*** 0.650 26.137*** 2.974
Switzerland 237 26.027*** 1.039 23.702*** 3.318
UnitedStates 1,270 24.440*** 0.903 42.316*** 3.034 +++
NorthAmerica 1,270 24.440*** 0.903 42.316*** 3.034 +++
LatinAmerica 68 24.050*** 0.638 65.337*** 1.008 +++
Europe 213 24.238*** 0.371 29.221*** 1.184 +++
Africa 13 5.954*** 1.990 54.123*** 4.408 +++
(2)%changeduring
othernations'matches
(3)%changeduring
ownmatches

Table3b:Volumetradedperminute,andpercentagechangeduringWorldCupmatches
(1)Volumepermin,
outsideWorldCup
Comparison
(2)vs.(3)
All 276,693 33.059*** 0.240 55.188*** 1.331 +++
Argentina 16,389 68.620*** 1.430 79.764*** 4.960 ++
Brazil 562,462 28.494*** 1.198 74.549*** 1.178 +++
Chile 321,560 79.030*** 1.183 99.513*** 0.397 +++
Denmark 8,098 30.951*** 2.250 6.145 8.984 +++
England 2,303,583 32.037*** 0.702 26.528*** 5.526
France 453,307 26.720*** 1.223 37.745*** 4.059 +++
Germany 159,641 36.975*** 0.964 59.519*** 2.000 +++
Italy 3,016,058 13.852*** 1.753 19.399*** 4.979
Mexico 410,283 24.358*** 1.284 63.298*** 2.281 +++
Netherlands 215,749 32.422*** 1.077 33.829*** 2.496
Portugal 122,888 47.333*** 1.811 56.985*** 3.266 ++
SouthAfrica 15,789 14.058*** 2.061 62.891*** 5.196 +++
Spain 594,593 28.892*** 0.994 30.687*** 3.387
Switzerland 129,836 37.535*** 1.182 39.972*** 4.658
UnitedStates 488,720 23.524*** 0.630 42.779*** 2.438 +++
NorthAmerica 488,720 23.524*** 0.630 42.779*** 2.438 +++
LatinAmerica 417,159 43.182*** 0.622 77.405*** 0.758 +++
Europe 163,045 31.571*** 0.400 37.745*** 1.456 +++
Africa 15,789 14.058*** 2.061 62.891*** 5.196 +++
(2)%changeduring
othernations'matches
(3)%changeduring
ownmatches

Note:Thetablesshowincolumn(1)themediannumberoftradesperminute(Table3a)andthetrading
volume (Table 3b), for time windows that correspond to match times, yet which lie outside the period
whentheWorldCupwasplayed.Thesewerecomputedfollowingequation1inthemaintext.Column(2)
showsthepercentagereductionintherespectivenumbersobservedduringmatchesbyothernations,and
column (3) the percentage change during matches by the national team. These figures correspond to
equation 2 in the main text. The last column tests whether the numbers in column (3) are significantly
lowerthanthoseincolumn(2).***/**/*and+++/++/+denotestatisticalsignificanceatthe1%/5%/10%
level.Allestimationsbasedonmedianregressions.

29
Table4a:Tradesperminute,andpercentagechangeduringWorldCupmatchesbycontinent
(A) (B) (C) (D) (A) (B) (C) (D)
(A)NorthAmerica 24.440*** 0.903 +++ 42.316*** 3.034 +++ +++ ++
(B)LatinAmerica 24.050*** 0.638 +++ 65.337*** 1.008 +++ +++ +++
(C)Europe 24.238*** 0.371 +++ 29.221*** 1.184 +++ +++ +++
(D)Africa 5.954*** 1.990 +++ +++ +++ 54.123*** 4.408 ++ +++ +++
comparisonversus comparisonversus
%changeduringownmatches %changeduringothernations'matches

Table4b:Volumetradedperminute,andpercentagechangeduringWorldCupmatchesby
continent
(A) (B) (C) (D) (A) (B) (C) (D)
(A)NorthAmerica 23.524*** 0.630 +++ +++ +++ 42.779*** 2.438 +++ + +++
(B)LatinAmerica 43.182*** 0.622 +++ +++ +++ 77.405*** 0.758 +++ +++ +++
(C)Europe 31.571*** 0.400 +++ +++ +++ 37.745*** 1.456 + +++ +++
(D)Africa 14.058*** 2.061 +++ +++ +++ 62.891*** 5.196 +++ +++ +++
comparisonversus comparisonversus
%changeduringownmatches %changeduringothernations'matches

Note: The tables show, separately for each continent, the percentage change in the median number of
trades per minute (Table 4a) and the volume traded per minute (Table 4b) observed during matches by
othernations(leftpanel)andduringownmatches(rightpanel).Thetablesalsoprovidetheresultofatest
whether the respective numbers are significantly different across continents. ***/**/* and +++/++/+
denotestatisticalsignificanceatthe1%/5%/10%level.

3
0

T
a
b
l
e

5
:

T
r
a
d
i
n
g

a
c
t
i
v
i
t
y

a
n
d

m
a
t
c
h

e
v
e
n
t
s
:

b
e
n
c
h
m
a
r
k

a
n
d

r
o
b
u
s
t
n
e
s
s

T
r
a
d
e
s
V
o
l
u
m
e
T
r
a
d
e
s
V
o
l
u
m
e
T
r
a
d
e
s
V
o
l
u
m
e
T
r
a
d
e
s
V
o
l
u
m
e
T
r
a
d
e
s
V
o
l
u
m
e
T
r
a
d
e
s
V
o
l
u
m
e

4
.
9
2
9
*
*

3
.
3
5
4

9
.
6
5
8
*
*
*

9
.
4
0
1
*
*

7
.
9
6
0
*
*

7
.
9
2
1

9
.
2
3
4
*
*
*

1
2
.
3
3
6
*
*

3
.
0
3
7

1
0
9
.
8
7
7

0
.
1
9
9
*
*

0
.
1
3
2
(
2
.
4
4
6
)
(
2
.
4
7
1
)
(
2
.
7
4
0
)
(
4
.
5
0
7
)
(
3
.
1
2
5
)
(
7
.
5
8
7
)
(
3
.
1
7
5
)
(
5
.
8
0
3
)
(
4
.
2
6
5
)
(
2
4
4
.
8
8
9
)
(
0
.
0
8
1
)
(
0
.
0
8
3
)
4
.
5
2
8
1
.
6
6
9

0
.
8
0
6

3
.
0
2
1
*
0
.
5
2
3

7
.
6
5
7
0
.
3
3
0

8
.
5
5
2
*
*

4
.
5
4
6

9
9
.
0
1
7

0
.
0
0
1

0
.
0
0
2
(
3
.
4
0
7
)
(
2
.
5
1
5
)
(
2
.
5
0
9
)
(
1
.
8
3
5
)
(
2
.
2
3
9
)
(
5
.
8
8
7
)
(
2
.
3
4
1
)
(
3
.
8
6
7
)
(
3
.
2
5
2
)
(
1
7
3
.
1
5
0
)
(
0
.
0
5
6
)
(
0
.
0
5
7
)
7
.
1
0
1
*
*
*
6
.
8
5
2
*
*
*
0
.
2
6
5
5
.
5
5
8
*
6
.
2
9
3
*
*
1
4
.
7
7
3
*
*
6
.
5
6
1
*
*
*
1
5
.
7
4
9
*

9
.
6
2
7
*
*
*

1
3
4
.
1
0
2
0
.
0
2
3
0
.
0
3
9
(
2
.
0
0
1
)
(
2
.
4
1
7
)
(
2
.
0
3
4
)
(
2
.
9
4
5
)
(
2
.
4
6
1
)
(
6
.
8
0
5
)
(
2
.
4
6
1
)
(
8
.
6
5
1
)
(
3
.
5
0
0
)
(
1
6
3
.
2
6
9
)
(
0
.
0
6
2
)
(
0
.
0
6
3
)
9
.
6
8
3
*
*
*
1
0
.
9
6
1
*
*
*

0
.
5
5
5

0
.
8
4
4
8
.
2
2
9
*
*
*
1
4
.
9
3
2
*
*
8
.
0
4
4
*
*
*
1
4
.
1
8
0

2
.
7
6
6

1
3
1
.
0
7
2
0
.
2
4
9
*
*
*
0
.
2
0
6
*
*
*
(
2
.
3
8
1
)
(
2
.
9
2
7
)
(
2
.
5
4
1
)
(
3
.
4
6
3
)
(
2
.
4
8
1
)
(
6
.
8
5
9
)
(
3
.
0
5
1
)
(
9
.
9
9
7
)
(
3
.
5
6
3
)
(
1
6
9
.
6
7
0
)
(
0
.
0
6
2
)
(
0
.
0
6
4
)
1
0
.
9
7
9
*
*
*
1
3
.
6
9
4
*
*

3
.
8
1
0

9
.
0
5
1
*
*
1
1
.
0
6
7
*
*
*
9
.
2
4
5
1
2
.
3
9
5
*
*
*
1
4
.
0
4
8
*
*
4
.
8
6
9

1
3
5
.
0
1
5
0
.
2
8
8
*
*
*
0
.
2
7
3
*
*
*
(
4
.
1
1
3
)
(
6
.
1
3
6
)
(
3
.
0
9
9
)
(
4
.
2
2
5
)
(
3
.
8
3
6
)
(
8
.
9
3
4
)
(
3
.
2
1
5
)
(
5
.
5
9
1
)
(
5
.
3
8
9
)
(
2
3
4
.
5
2
9
)
(
0
.
0
9
0
)
(
0
.
0
9
5
)
1
2
.
5
9
2
*
*
*
1
3
.
4
3
3
*
*
*
3
.
0
6
2

3
.
1
5
4
1
0
.
5
0
3
*
*
*
7
.
8
8
8
1
1
.
8
2
9
*
*
*
1
2
.
6
8
0
*
*
6
.
7
6
7

1
3
1
.
6
5
1
0
.
1
1
7
0
.
1
8
1
*
*
(
4
.
5
2
6
)
(
4
.
9
5
2
)
(
3
.
0
5
2
)
(
4
.
0
1
9
)
(
3
.
8
4
0
)
(
8
.
9
3
5
)
(
3
.
2
9
8
)
(
5
.
6
0
0
)
(
5
.
3
9
4
)
(
2
3
3
.
7
8
9
)
(
0
.
0
8
7
)
(
0
.
0
9
1
)
2
7
.
3
7
1
*
*
*
3
6
.
1
0
5
*
*
*
4
.
7
4
8
1
.
2
6
9
2
2
.
1
9
0
*
*
*
4
0
.
8
1
1
*
*
*
2
3
.
5
1
6
*
*
*
4
5
.
6
0
4
*
*
*

1
1
.
9
4
4
*
*

1
0
7
.
2
6
5
0
.
2
3
7
*
*
*
0
.
2
3
1
*
*
(
4
.
4
3
9
)
(
5
.
8
6
7
)
(
3
.
8
8
9
)
(
7
.
6
4
0
)
(
3
.
8
4
0
)
(
8
.
9
3
5
)
(
3
.
0
9
5
)
(
1
3
.
8
7
2
)
(
5
.
3
9
4
)
(
2
3
3
.
7
8
9
)
(
0
.
0
8
7
)
(
0
.
0
9
1
)

4
7
.
3
0
5
*
*
*

5
6
.
7
5
5
*
*
*

2
7
.
3
4
5
*
*
*

2
9
.
0
9
3
*
*
*

3
7
.
8
0
7
*
*
*

3
5
.
4
4
4
*
*
*

3
9
.
1
3
3
*
*
*

4
0
.
2
3
6
*
*
*

0
.
9
3
0
1
2
9
.
1
2
2
*

0
.
1
6
6
*
*
*

0
.
1
7
6
*
*
*
(
1
.
1
7
9
)
(
1
.
1
6
9
)
(
1
.
1
6
7
)
(
1
.
1
4
0
)
(
1
.
1
1
6
)
(
3
.
1
2
1
)
(
1
.
1
9
0
)
(
2
.
5
9
1
)
(
1
.
6
1
4
)
(
7
1
.
8
9
7
)
(
0
.
0
2
8
)
(
0
.
0
2
9
)
O
b
s
e
r
v
a
t
i
o
n
s
3
,
7
2
1
3
,
7
2
0
3
,
4
8
7
3
,
4
8
7
3
,
7
2
1
3
,
7
2
0
3
,
7
2
1
3
,
7
2
0
3
,
6
3
8
3
,
6
3
8
3
,
2
9
0
3
,
2
9
0
R
2
0
.
0
1
8
0
.
0
0
8
0
.
0
1
2
0
.
0
0
9
S
h
i
f
t
e
d

t
i
m
e

s
c
h
e
d
u
l
e
A
l
t
e
r
n
a
t
i
v
e

d
e
f
i
n
i
t
i
o
n

o
f

a
b
n
o
r
m
a
l

t
r
a
d
i
n
g
B
e
n
c
h
m
a
r
k
O
L
S

r
e
g
r
e
s
s
i
o
n
Q
u
o
t
e
s
T
o
b
i
t

r
e
g
r
e
s
s
i
o
n
1
5

m
i
n
u
t
e
s

a
f
t
e
r

m
a
t
c
h
1
6

3
0

m
i
n
u
t
e
s

a
f
t
e
r

m
a
t
c
h
3
1

4
5

m
i
n
u
t
e
s

a
f
t
e
r

m
a
t
c
h
C
o
n
s
t
a
n
t
G
o
a
l
s
R
e
d

o
r

y
e
l
l
o
w

c
a
r
d
s
N
a
t
i
o
n
a
l

a
n
t
h
e
m
s
H
a
l
f

t
i
m
e

N
o
t
e
:

T
h
e

t
a
b
l
e

s
h
o
w
s

t
h
e

r
e
s
u
l
t
s

o
f

r
e
g
r
e
s
s
i
o
n

m
o
d
e
l

(
3
)
,

w
h
i
c
h

e
s
t
i
m
a
t
e
s

t
r
a
d
i
n
g

a
c
t
i
v
i
t
y

d
u
r
i
n
g
,

b
e
f
o
r
e

a
n
d

a
f
t
e
r

o
w
n

W
o
r
l
d

C
u
p

m
a
t
c
h
e
s

a
n
d

i
s

(
f
o
r

t
h
e

e
x
a
m
p
l
e

o
f

t
r
a
d
e
s

p
e
r

m
i
n
u
t
e
)

s
p
e
c
i
f
i
e
d

a
s

,
1
2
,
3
,
4
,
5
,
6
,
7
,
8
,
,
_
1
_
2
_
3
c
t
c
t
c
t
c
t
c
t
c
t
c
t
c
t
c
t
t
r
a
d
e
s
g
o
a
l
c
a
r
d
a
n
t
h
e
m
s
h
a
l
f
t
i
m
e
a
f
t
e
r
m
a
t
h
a
f
t
e
r
m
a
t
h
a
f
t
e
r
m
a
t
h
u
o
o
o
o
o
o
o
o
=
+
+
+
+
+
+
+
+
F
o
r

d
e
f
i
n
i
t
i
o
n

o
f

r
e
g
r
e
s
s
o
r
s
,

p
l
e
a
s
e

s
e
e

t
h
e

m
a
i
n

t
e
x
t
.

P
a
n
e
l

B
e
n
c
h
m
a
r
k

r
e
p
o
r
t
s

b
e
n
c
h
m
a
r
k

r
e
s
u
l
t
s
,

b
a
s
e
d

o
n

m
e
d
i
a
n

r
e
g
r
e
s
s
i
o
n
s
.

O
t
h
e
r

c
o
l
u
m
n
s

s
h
o
w

a
d
d
i
t
i
o
n
a
l

r
e
s
u
l
t
s
,

n
a
m
e
l
y

r
e
p
l
a
c
i
n
g

t
r
a
d
e
s

b
y

q
u
o
t
e
s

(
p
a
n
e
l

Q
u
o
t
e
s

)
,

c
o
n
d
u
c
t
i
n
g

l
e
a
s
t

s
q
u
a
r
e
s

r
e
g
r
e
s
s
i
o
n
s

(
p
a
n
e
l

O
L
S

r
e
g
r
e
s
s
i
o
n

)
,

t
o
b
i
t

r
e
g
r
e
s
s
i
o
n
s

(
p
a
n
e
l

T
o
b
i
t

r
e
g
r
e
s
s
i
o
n

)
,

a
r
t
i
f
i
c
i
a
l
l
y

s
h
i
f
t
i
n
g

b
a
c
k
w
a
r
d

t
h
e

t
i
m
i
n
g

o
f

a
l
l

m
a
t
c
h
e
s

b
y

1
2
0

m
i
n
u
t
e
s

(
p
a
n
e
l

S
h
i
f
t
e
d

t
i
m
e

s
c
h
e
d
u
l
e

)
,

a
n
d

u
s
i
n
g

a
n

a
l
t
e
r
n
a
t
i
v
e

d
e
f
i
n
i
t
i
o
n

o
f

a
b
n
o
r
m
a
l

t
r
a
d
i
n
g

(
p
a
n
e
l

A
l
t
e
r
n
a
t
i
v
e

d
e
f
i
n
i
t
i
o
n

o
f

a
b
n
o
r
m
a
l

t
r
a
d
i
n
g

)
.

*
*
*
/
*
*
/
*

d
e
n
o
t
e

s
t
a
t
i
s
t
i
c
a
l

s
i
g
n
i
f
i
c
a
n
c
e

a
t

t
h
e

1
%
/
5
%
/
1
0
%

l
e
v
e
l
.

N
u
m
b
e
r
s

i
n

i
t
a
l
i
c
s

a
r
e

r
o
b
u
s
t

s
t
a
n
d
a
r
d

e
r
r
o
r
s
.

31
Table6:Lesscomovementwithglobalstockmarketsduringnationalteammatches
Obs. R2
0.028 *** 0.866 *** 0.181 *** 442,118 0.235
0.002 0.002 0.029
0.327 *** 0.031 * 0.071 21,240 0.056
0.040
0.016 0.075
0.280 *** 0.669 *** 0.029 26,145 0.331
0.025
0.027 0.067
0.257 0.057 *** 0.102 ** 24,570 0.029
0.190
0.012 0.045
0.054 *** 0.791 *** 0.132 29,450 0.190
0.018
0.085 0.272
0.085 *** 1.087 *** 0.310 *** 32,130 0.416
0.013
0.067 0.077
0.056 *** 1.366 *** 0.375 *** 33,150 0.382
0.015
0.084 0.141
0.074 *** 0.995 *** 1.056 *** 33,150 0.305
0.019
0.050 0.209
0.016 1.532 *** 0.008 32,320 0.359
0.014
0.105 0.149
0.139 *** 0.465 *** 0.235 *** 25,343 0.220
0.030
0.067 0.087
0.063 *** 1.172 *** 0.155 33,150 0.353
0.013
0.077 0.104
0.074 *** 0.561 *** 0.175 ** 32,640 0.082
0.019
0.062 0.084
0.054 *** 0.798 *** 0.378 ** 29,610 0.214
0.019
0.070 0.149
0.007 1.695 *** 0.236 33,150 0.344
0.013
0.096 0.150
0.087 *** 0.655 *** 0.021 31,500 0.197
0.011
0.037 0.077
0.104 ** 1.295 *** 0.287 *** 24,570 0.709
0.048 0.028 0.106
Switzerland
Spain
UnitedStates
SouthAfrica
Germany
Denmark
France
Netherlands
Portugal
Argentina
Brazil
England
Mexico
Chile
Italy
Decoupling Beta Ownlags
All

Note:Thetableshowstheresultsofregressionmodel(4),whichestimatesnationalstockmarketreturns
asafunctionofownlags,globalstockmarketreturns,dayoftheweekeffects,timeofthedayeffects,time
ofownmatches,andaninteractionofglobalstockmarketreturnsandthetimingofownmatches,andis
specifiedas
, 1 2 , 1 3 , 2 4 , 3 , , , 1 , 2 , 3 , , c t c t c t c t w t w t c t c t c t c t c t
r r r r r r match match dow tod u o o o o | o

= + + + + + + + + + ColumnOwn
lags reports the sum of coefficients on the own lags, Column Beta the coefficient on the global stock
market returns (|), and Column Decoupling thecoefficient onthe interactionterm (o). ***/**/* denote
statistical significance at the 1%/5%/10% level. For the pooled results, numbers in italics are panel
corrected standard errors which allow the variance of u
c,t
to be countryspecific, but do not allow for
contemporaneouscrosscountrycorrelation.Forthecountryresults,numbersinitalicsareWhite(1980)
robusterrors.

32
Table7:Robustnesstestsforoverallchangeintradingactivityduringownmatches
(1)Activitypermin,
outsideWorldCup
Comparison
(2)vs.(3)
Numberofquotes 787 16.289 *** 0.286 28.39 *** 1.046 +++
Volumequoted 5,706,251 14.469 *** 0.386 30.10 *** 0.573 +++
Numberoftrades 381 12.897 *** 6.224 37.148 *** 6.340 +++
Volumetraded 610,382 3.139 0.379 35.291 *** 6.316 +++
Numberoftrades 381 13.226 *** 0.315 38.758 *** 0.919 +++
Volumetraded 610,382 7.859 *** 1.666 40.474 *** 1.687 +++
Numberoftrades 24 28.780 *** 0.433 n.a.
Volumetraded 1,491 47.940 *** 0.917 n.a.
Numberoftrades 222 7.814 4.848 3.857 5.430
Volumetraded 355,598 1.079 5.763 0.024 6.788
Numberoftrades 0 0.018 *** 0.006 0.144 *** 0.023 +++
Volumetraded 0 0.009 0.006 0.156 *** 0.023 +++
Alternativedefinition
ofabnormaltrading
Shiftedtimeschedule
Quotes
OLSregression
Germangovernment
bondfutures
Tobitregression
(2)%changeduringother
nations'matches
(3)%changeduringown
matches

Note:ThetableshowstheresultsofasetofrobustnessteststoTable3,bytestingwhetherthepercentage
changeinthemediannumberoftradesperminuteandthevolumetradedperminuteobservedduring
matchesinwhichthenationalteamcompetedisobservedalsowhenreplacingtradesbyquotesgivenby
marketmakers(panelQuotes),whenconductingleastsquaresregressionsratherthanmedian
regressions(panelOLSregression),whenusingatobitregressiontoaccountfortheleftcensoringofthe
dependentvariable(panelTobitregression),whenusingdataonGermangovernmentbondfuturesfor
allGermansoccercupmatchesduringtradingtimefrom19982010(panelGermangovernmentbond
futures),whenartificiallyshiftingbackwardtimingofallmatchesby120minutes(panelShiftedtime
schedule),andwhenusinganalternativefilteringschemeasinEdmansetal.(2007)andGallantetal.
(1992)toconstructmeanzero,unitvarianceseriesforabnormaltradesortradingvolumes(panel
Alternativedefinitionofabnormaltrading).***/**/*and+++/++/+denotestatisticalsignificanceatthe
1%/5%/10%level.Numbersinitalicsarerobuststandarderrors.Incaseofquotesandtheshiftedtime
schedule,thesestandarderrorsaccountforclusteringbycountry.

33
Table8:Decoupling:differentcomovementduringlargeandregularmovementsofglobalindex
Ownlags 0.028 *** 0.002 0.028 *** 0.002
Betaforregularmovements 0.943 *** 0.010 1.014 *** 0.014
Decouplingforregularmovements 0.395 *** 0.060 0.455 *** 0.085
Betaforlargemovements 0.862 *** 0.002 0.862 *** 0.002
Decouplingforlargemovements 0.130 *** 0.034 0.159 *** 0.031
Obs.
R2
442,118
0.235
(1)
442,118
0.235
(2)

Note: See table 6. The results are based on a regression model that further differentiates whether the
movements in global stock markets are large or regularsized. In column (1), global stock market
movements are classified as large if they are above the 90
th
or below the 10
th
percentile of the sample
distribution.Incolumn(2),largeglobalstockmarketmovementsarethosebeyondthe85
th
orbelowthe
15
th
percentile of the sample distribution. ***/**/* denote statistical significance at the 1%/5%/10%
level.Numbersinitalicsarepanelcorrectedstandarderrorswhichallowthevarianceofu
c,t
tobecountry
specific,butdonotallowforcontemporaneouscrosscountrycorrelation.

34
Table9:Comparingdecouplingduringsoccermatchesandlunchhours
Obs. R2
0.028 *** 0.868 *** 0.182 *** 0.018 ** 442,118 0.235
0.002 0.002 0.029 0.008
Decoupling
duringmatches
Ownlags Beta Decoupling
duringlunch
All

Note:Thetableshowstheresultsofregressionmodel(5),whichestimatesnationalstockmarketreturns
asafunctionofownlags,globalstockmarketreturns,dayoftheweekeffects,timeofthedayeffects,time
ofownmatches,lunchtime,aninteractionofglobalstockmarketreturnsandthetimingofownmatches,
and an interaction of global stock market returns and lunchtime, and is specified as
, 1 2 , 1 3 , 2 4 , 3 , 1 , , 2 , , 1 , 2 , 3 , 4 , , c t c t c t c t w t w t c t w t c t c t c t c t c t c t
r r r r r r match r lunch match dow tod lunch u o o o o | o o

= + + + + + + + + + + +
Column Own lags reports the sum of coefficients on the own lags, Column Beta the coefficient on the
global stock market returns (|), Column Decoupling during matches the coefficient on the first
interaction term (o
1
), and , Column Decoupling during lunch the coefficient on the second interaction
term (o
2
). ***/**/* denote statistical significance at the 1%/5%/10% level. Numbers in italics are panel
corrected standard errors which allow the variance of u
c,t
to be countryspecific, but do not allow for
contemporaneouscrosscountrycorrelation.

35
Table10:Resultsfordispersionofstockreturns
(1)Activitypermin,
outsideWorldCup
Comparison
(2)vs.(3)
Returndispersion 0.114 19.595 *** 0.247 21.596 *** 0.879 +++
(2)%changeduringother
nations'matches
(3)%changeduringown
matches

Note:Seetable3.Thistablepresentadditionalresultsforthestandarddevationofallindividualstockina
countrysindex.***/**/*and+++/++/+denotestatisticalsignificanceatthe1%/5%/10%level.
Estimationsbasedonmedianregressions.