International Football Results and Stock Returns Evidence for the World Cup and European Championship

(1)

International Football Results and Stock Returns

Evidence for the World Cup and European Championship

Master Thesis

MSc Business Administration – Finance

Risk and Portfolio Management

By

Lútzen H. Brink

S1383981

Student at the University of Groningen Faculty of Economics and Business

21 may 2009

Supervisor: dr. T.T.T. Pham

Faculty of Economics and Business University of Groningen

Second Supervisor: drs. M.M. Kramer

(2)

International Football Results and Stock Returns

Evidence for the World Cup and European Championship

Lútzen H. Brink S1383981 lutzenbrink@gmail.com

JEL codes: A12, G14, C22

Key words: Football, Investor Mood, Goals, Stock Market Reaction

Abstract:

Psychological evidence suggests that the outcome of a football match causes mood swings, which influences investors’ sentiment. This thesis examines the relationship between the outcome of an international football match and the stock market reaction, the trading day after the match, at the country’s leading stock exchange, from 1974 to 2008. The main finding of this thesis is that there is a loss effect at the World Cup that leads to an abnormal return of -0,22%. The strongest loss effect, measured at the World Cup, is a loss in which the national team conceives a small amount of goals, 0 goals or 1 goal, that leads to an abnormal return of -0,47%.

I. Introduction

(3)

Market participants can regard the result of an international football match as information. For example, translating the a winning outcome of an international football match in to higher revenues for beer breweries. This outcome of the match or new information can be associated with the financial consequence in the sense of income and might change the perception of valuation. An alternative perspective to look at this phenomenon is that an international football match affects the mood of people. Mood is one of the behavioural biases which affects the decision making and judgements of people (Hirshleifer, 2001). Due to these behavioural biases people tend to make irrational decisions. In the psychological literature, extended in the next session, it is argued that people tend to misattribute their mood to different or wrong features of the environment. For example, when the national football team wins, people are in a good mood, this may cause people to be unconsciously more optimistic when evaluating propositions. When this misattribution stretches out to investments, stock prices will fluctuate with the mood swings of investors (Nofsinger, 2005). Therefore, when the national football team loses it is hypothesized that it will result in a stock market decline, whereas victories of the national football team will result in appreciation of the stock market.

To test the efficient market hypothesis (Fama, 1970) two different models are used. The main model is the regression model of Edmans et al. (2007). The second method is the event study model of Brown and Warner (1985) and MacKinlay (1997). This thesis examine whether there is a stock market reaction on the country’s leading stock exchange, the trading day after an international football match, from 1974 to 2008. Whereby, 27 countries are being included in the dataset. The international football matches that are considered as important are the matches at the World Cup and European Championship.

(4)

This thesis is inspired by study of Edmans et al. (2007). They use the “novel mood variable, international soccer results, to investigate the effect of investor sentiment on asset prices” (Edmans et al., p1, 2007). They test 39 national team results on World Cups and Continental Cups over the period 1973 to 2004. They find a loss effect the day after an international football match is played of -0,21%. The largest significant loss effect is found during the World Cups. The abnormal return the trading day after a football match on a World Cup is -0,49%. This thesis extends the research performed by Edmans et al. (2007). First, it makes solely the distinction between the international football matches played at World Cups and European Championships. Second, this thesis test whether the win or loss effect found by Edmans et al. (2007) can be attributed to the amount of goals scored or the difference in goals scored in an international football match. Third, this thesis tests whether the results are time-period dependant and if markets have become (in)efficient over the years. Fourth, this thesis tests whether the loss or win effect found by Edmans et al. (2007) can be attributed to a certain group of nations on the World Cup or on the European Championship.

The rest of this paper is organized as follows. Section II describes how mood is misattributed to international football matches and stock returns. Section III contains the methodology. Section IV describes the data that is used for the analysis. Section V gives the results and interpretates these results. Section VI contains the discussion of the research and Section VII concludes.

II. Literature review

2.1 Mood, Judgement and Decision Making

(5)

information processing errors such as representativeness and framing. Emotion is the behavioural bias that is affected by mood and herding is an example of social interaction.

Moods have profound impact on memory, judgement, perception and behaviour. According to Nofsinger (2005) moods can influence decision making. In decision making people tend to underestimate the influence of these mood swings (Nofsinger, 2005). When you are happy everything looks fantastic and when you are sad the world looks in gloom and doom, this is the so called mood-congruency effect (Avramova and Stapel, 2008). People who are subject to good moods, have a tendency to evaluate many kinds of activities more positive (Wright and Bower, 1992; Clore et al., 1994). On the other hand, people who are subject to bad moods find negative

information more salient (Isen, 1984; Isen et al, 1978). Isen et al .(1978)1_{makes a congruent}

relationship between positive and negative moods, if people have a negative mood they will look for negative confirmation material and, the other way around, people having a good mood look for positive confirmation material. The magnitude of the role of moods will strongly affect judgement if people, having a good mood or lack concrete information (Clore, Schwarz and Conway, 1994).

Several studies found that people having a good mood are handling information less critical, relying on the automatic responses generated by our heuristics. Whereas people having a bad mood are handling information more in depth and in a diagnostical way (Schwarz, 1990). People with good moods tend also to be more open to other opinions, no matter if it is a good argument or a bad argument (Mackie and Worth, 1991). This leaves judgements vulnerable to erroneous interpretations. Considering the various positive and negative attributes of an object to make a final judgement, people make use of the so called simplifying heuristics (Schwarz and Clore, 1983). The judgement is based on the feeling you have about an object. People who are having a positive or a negative mood may mistake their feeling towards an object in formulating a judgement. This may cause misattribution, attributing feelings to the wrong source in the environment, leading to incorrect judgements. Furthermore, people do not draw all the relevant information from their memory, but base their judgement on the information that is most

(6)

accessible (Wyer & Srull, 1986). Because people do not make a judgement based on the all the information, people do not make optimal decisions. People make decisions that satisfy. Whereby the satisfying choice, under the conditions of bounded rationality, causes mood-congruented information to be over weighted in the interpretation of an object and consequently in the judgement of it.

For example, with theory of Schwarz and Clore (1983) about the problem of misattribution in mind, people are happier when their football team wins than when their football team looses. Moreoverr, if you first ask them about the outcome of the football match, than their judgement about their own state of happiness reduces. This would hypothesize that if you separate the environment from the long term consideration, people will attribute their state of happiness to the outcome of the football match instead of a long term consideration. As a consequence, the judgements are influenced in a mood-congruent manner (Schwarz and Clore, 1983).

2.2 Mood, Football Results and Behaviour

Psychologist have layed a foundation for the relationship between football result and behaviour.

Some of the investigations link football matches to suicide2_{(Trovato, 1998), riots (Wann et al.,}

2001) or heart attacks3_{(Carroll et al., 2002; Berthier and Boulay 2003; Chi and Kloner 2003).}

Nofsinger (2008) argues that mood influences investors’ decisions and so he argues that the misattribution extends into the financial domain of setting prices. This means that if the nations’ football team wins, people are more inclined in buying stocks at the country’s leading stock exchange (Edmans et al., 2007). Thus, incorrectly attributing their good mood to stock prices rather than attributing their good mood to the win of the nations’ football team. On the other hand, when the nations’ football team looses, people are more inclined to sell stocks on the nations’ stock exchange (Nofsinger, 2008). Mood influences the buying and selling behaviour, because investors are making educated guesses for the assumptions in the quantitative methods in fundamental analysis. Where optimism leads to higher forecast assumptions and pessimism leads to lower forecast assumptions (Nofsinger, 2005).

2_{Fans can even be so disappointed by a defeat of their favourite team that they become suicidal}

(7)

The prediction that a certain nation will win a football match, as with the weather forecast that tomorrow will be sunny, is not immediately reflected in the stock prices. Furthermore, people posses a form of allegiance bias, which causes people to think that their own team has a bigger chance of winning (Markman and Hirt, 2002). Rather, it is hypothesized that after the occurrence of the football game there should be a change in stock price observed at the country’s leading stock exchange.

If people are rational, under the condition of unbounded rationality, than there is no reason to assume that stock prices are correlated with the outcome of international football matches. Therefore, to rationalize the link between the stock price change of national stock exchange and a mood swing among the citizen of that same country, the event must match, according Edmans et al. (2007), a couple of criteria:

1) An event should drive mood of a population in such a way that it can impact the returns on the country’s leading stock exchange.

2) An event should be noticed by a large group of national citizens, in such a way that the results can be known to a large part of the population.

3) The result should be interpretated the same, among the majority of all the citizens.

Hirt et al. (1992)4_{examined the first criteria in which the relationship is tested between fanship of}

a team and the predictive evaluative behaviour after an American Football match is played. For

the second criteria, it is argued that football is watched by billions of people5_{and a very high}

percentage of up to 80% of the nation watched an international football game on a World Cup

and European Championship6_{. For the third criteria, Edmans et al. (2007) argue that football is of}

“national interest” (Edmans et al., p2, 2007). Furthermore, people invest money far more in their

4_{Hirt et al. (1992) made two different groups of students. One group were the fans of the winning team and the other} group were the fans of the loosing team. After the match, the individuals in both groups were asked to give an estimate of their own future performance. The winning group estimated their own performance significantly higher than the loosing group. The group whose team won gave a more optimistic own future performance. On the other hand, the loosing group was more pessimistic about their own future performance.

5_{In 2002 the World Cup had, in cumulative terms, 26,4 billion viewers around the world and in 2006 in cumulative} terms 26,29 billion viewers around the world. The European Championship had, in comparison, 7,8 billion viewers in cumulative terms

6_{During the European Football Championship in Austria and Switzerland, in 2008, the football matches grabbed a} television market share of over 80% in Portugal, Netherlands and Germany. 70% of the Polish and Swiss people watched their own nation play. In Italy, Spain, Greece and Portugal for the public television broadcaster it is even mandatory to broadcast the football matches live.

(8)

own country than portfolio theory would suggest (French and Poterba, 1991; Huberman, 2001). French and Poterba (1991) showed that on average 90% of the countries stock market is owned by domestic investors. To sum up, international football matches played at the World Cup and the European Championship are showing the three characteristics outlined by Edmans et al. (2007), that rationalize the link between stock returns and international football matches. Football creates, what Edmans et al. (2007) call, a large signal-to-noise ratio.

2.3 Mood, International Football Matches and Stock Prices

The behavioural finance literature has extended their research with investigation about the influence of sunshine (Saunders, 1993), lunar cycles (Yuan et al., 2005), daylight (Kamstra et al., 2003), football (Edmans et al., 2007), temperature (Cao and Wei, 2005) and rugby (Boyle and Walter, 2002) on the mood of investor and subsequently, being the novel variable for the announcement effect (over a very short period of time). These studies show a link between mood and the stock market reaction on the country’s leading stock exchange the trading day after the event. Consequently, investors are influenced by their mood in their decision making and cause markets to set different prices after the event.

For public traded football clubs the impact of a match can be measured in terms of both wins and losses. Also the market efficiency of betting quotations has been examined (Forrest et al., 2005; Palomino et al., 2005.) So far more and more research has been done to examine the results of football clubs and the effect on the underlying stock value of the football club (Zuber, 2005; Peenstra and Scholtens, 2008; Ashton et al., 2003; Renneboog and Vanbrabant, 2000; Stadtmann, 2006; Palomino, 2005).

(9)

loss in a football match gives a drop of -0,38% at the country’s leading stock exchange, the trading day after the football match took place, up to -0,49% if a loss occurs in the World Cup elimination phase. Edmans et al. (2007) conclude that football matches influence the buy and sell behaviour of these investors and traders.

In this thesis the relationship is examined between stock returns and the mood variable international football matches. In this thesis, 1) the hypothesis that will be tested is that a win will correlate positively to the stock return of the nation’s stock exchange and a loss will correlate negatively to the stock return of the nation’s stock exchange. Furthermore, psychological literature suggest that a loss is perceived differently in comparison to a win. The impact of a loss in a football tournament should have a bigger impact because of the fact that after a loss, an exit will follow from the competition format. This is similar to the prospect theory of Kahneman and Tversky (1979). According to the prospect theory of Kahneman and Tversky (1979), in a football

match a win and a loss are not perceived as a symmetric event7_{. Therefore, it is important to make}

a distinction between a perceived loss and a perceived win. So, 2) it is hypothesized that a losing match should make a bigger impact than a winning match on the stock return of the nation’s national stock exchange.

Even if an international football match has a relationship with the returns at country’s leading stock exchange the trading day after the event, there can be more underlying factors during an international football match which can explain the previous mentioned relationship. In a football match, the outcome of a football match is derived from the number of goals scored. Zuber (2005) finds no significant results for English public traded football teams. The difference in goals scored in football matches of English public traded football show no significant effects on the stock returns of the English public traded football teams. However, 3) it can be hypothesized that

the amount8_{of goals scored or difference}9_{in goals, in both winning and losing matches, will}

have an impact on the change of stock market prices at the country’s leading stock exchange.

7_{According to the prospect theory a defeat can be two and a half times as painful as the feeling towards expected} win.

(10)

III. Methodology

3.1 Empirical Design

To examine the hypothesis empirically the regression model of Edmans et al. (2007) is used. As a comparison the event study methodology of Brown and Warner (1985) and MacKinlay (1997) is used. Furthermore, robustness checks are performed.

3.2 The regression model

The regression model of Edmans et al. (2007) is considered as the main method. If an event study approach is followed, one can conclude whether wins and losses result in variation in stock performance. However, one cannot tell how much of the variation in stock returns can be attributed to a win or a loss. Therefore, the regression model of Edmans is considered as the main method. The response of the market for both the game winner and the game loser can be measured by, first, estimating the following model for each country i (Edmans et al., 2007):

it mt i

it R

R =γ0 +γ1 +ε ( 1 )

In which Rit is the daily total return index, measured in local currency of the stock exchange of

country i on the first trading day t after the event.Rmtdenotes the daily MSCI World total return

index on the first trading t after the event. And εit is the error term. The nations national stock

exchange will be correlated with the world market index. While for some football matches the matches are played on a weekday. The trading day after the match ensures that a full trading is taken into account.

Second, with the residuals from equation (1) we can estimate the effect of the wins and losses of a international football match. Edmans et al. (2007) use the regression model:

(11)

In which εˆit are the residuals from equation (11). Wit denotes the dummy variable for wins that

takes the value 1 if country i wins and the value 0 otherwise on the first trading day t after the

event, with βW as the corresponding coefficient. Litdenotes the dummy variable for losses that

takes the value 1 if country i loses and the value 0 otherwise, with βLas the corresponding

coefficient . And, υitis the error term.

The results on the Ordinary Least Squares (OLS) method assumes that the variance and error terms are constant. Therefore, it needs to be tested for autocorrelation, looking at the Durbin Watson statistic, to look if the error terms are uncorrelated. With the White’s heteroskedasticity Test (1980) it can be measured whether the residual variance of a variable in a regression model is constant. If this is the case, the regression should be modelled using the autoregressive conditional heteroskedasticity (ARCH) (Engle, 1982) model to see if there are ARCH effects. If the assumptions of the OLS are violated and ARCH effects occur, a GARCH (1,1) model is used:

∑

= + = − + = − q i p i t i t t 1 1 2 ) 1 ( 2 0 2 ) 1 ( βσ ε α α σ ( 3 )

3.2.1 Extending the regression

Besides the effect of the wins and the losses on World Cups and European Championships, an extended model is used to measure the relationship between the stock market returns at the country’s leading stock exchange and other (mood) variables.

First, the effect of the difference in goals scored or conceived in an international football match. This can be calculated with regressing the residuals of equation (1):

it it L it L it it W it W it it W β RS β RB L β RS β RB υ εˆ = ( ₁* + ₂* )+ ( ₁* + ₂* )+ _{( 4 )}

In which εˆitare the residuals from equation (1). Wit denotes the dummy variable for wins that

(12)

event, with βW1 and βW2as the corresponding coefficient dependant on the variables RSit and it

RB . RSit and RBit are defined as the difference in goals scored or conceived in a match.

it

RS is defined as the small difference in goals scored, being 010_{or 1.} it

RB is defined by the big

difference in goals scored, scoring or conceiving 2 or more goals.Litdenotes the dummy variable

for losses that takes the value 1 if country i loses and the value 0 otherwise, with βL1andβL2 as

the corresponding coefficient dependant on the variables RSit and RBit. And, υit is the error

term. Four subsamples are created to keep a large number of observations in every subsample. For the amount of goals scored or conceived in an international match, the same equation (4) can be used. However, details change in the equation. So, regressing the residuals of equation (1):

it it L it L it it W it W it

it W β AGSS β AGSB L β AGCS β AGCB υ

εˆ = ( ₁* + ₂* )+ ( ₁* + ₂* )+ _{( 5 )}

In which εˆit are the residuals from equation (1). Wit denotes the dummy variable for wins that

takes the value 1 if country i wins and the value 0 otherwise on the first trading day t after the

event, with βW1 and βW2as the corresponding coefficient dependant on the variables AGSSit

and AGSBit. AGSSit and AGSBit are defined as the amount of goals scored in a match,

separated into two groups. AGSSit is a small amount of goals scored, being 0 or 1. AGSBit a

big amount of goals scored, being 2 or more goals. Lit denotes the dummy variable for losses

that takes the value 1 if country i loses and the value 0 otherwise, with βL1andβL2 as the

corresponding coefficient dependant on the variables AGCSit and AGCBit. AGCSit and

it

AGCB _{are defined as the amount goals conceived.}AGCSit is a small amount of goals

conceived, being 0 or 1. AGCBit is a big amount of goals conceived, being 2 or more goals.

And, υitis the error term.

(13)

3.3 The event study method

MacKinley (1997) argues that the abnormal return can be computed by the normal return on the trading day after the event of the country’s leading stock exchange of the event window minus the normal return of the index in the event window. Whereas the normal return is the expected return which considers that the football match did not take place.

) | ( it t it it R E R X AR = − ( 6 )

In which ARit denotes the abnormal return for the index i and t the first full day of trading after

the football match. Rit is the stock market reaction the trading day after the match at the

country’s leading stock exchange.E(Ri_τ |X_τ)denotes the expected normal return for the

estimation period

τ

. For the estimation period the approach by Brown and Warner (1985) is

used. Therefore, an event estimation window of 250 days will be used, ending at 30 days prior to the event. A championship has a maximum duration of 1 month. Hence, the estimation period is not affected by an event-related return of a football team. The event period is 1 single day, the first trading day after a football match. The 1 single day event period is taken, because of the fact that during the World Cup and European Championship matches are played within at least 3 days of each other. Therefore, event returns will not overlap each other. Glascock et al. (1991) point out that taking a 1 single day event period will result in a more direct and uncontaminated result of the event, also performed by Scholtens and Peenstra (2008).

.

3.3.1 The return model

The actual ex post return of the index is stated as:

(14)

In which Rit denotes the stock market reaction the trading day t after the match at the leading

stock exchange of country i. Pit denotes the index quotation of country i on the first trading day

t after the football match. In which ζitdenotes the time period t error term for index i with an

expectation of zero and a varianceσ_ζ2_i .

3.3.2 The market model

Assuming that stock market returns are normally distributed (Brown and Warner, 1985), the one-factor OLS market model will be used:

. ₍ ₎ ₀ _var( ₎ 2 i it it it mt i i it E R R ε σ ε ε ε β α = = + + = ( 8 )

In which Rit denotes the stock market reaction the trading day t after the match at the leading

stock exchange of country i and Rmt denotes the return of the MSCI World total return index m

of the market portfolio over period t. εit is the error term. αi,βi and 2

i

ε

σ are the parameters of

the one-factor OLS market model. In which, ) ( ) , ( mt mt it i R VAR R R Cov = β _{( 9 )}

3.3.3 The statistical properties of the abnormal return

With these market model parameters we can measure and analyze the abnormal returns for index

i in the event window. MacKinlay (1997) states that the abnormal market return becomes:

mt i i it it R R R Aˆ = −αˆ −βˆ ( 10 )

The abnormal return is derived from the value of the dependant variable and what the model

predicts from the alpha and the beta . The conditional variance will be 2(ˆ )

it

R A

(15)

          ₋ + + = _∧ 2 2 2 2₍ˆ ₎ ₁ ( ˆ ) m m mt it R R A i σ µ σ σ ε . ( 11 )

3.3.4 The aggregation of the abnormal returns

The abnormal return observations must be aggregated in order to draw a conclusion for the impact of the outcome of football matches at the country’s leading stock exchange (MacKinlay, 1997). The aggregation is among the indices. MacKinlay (1997) makes two assumptions. First, there is no clustering involved. Second, the abnormal index returns are independent.

The aggregation of the abnormal index returns can be done using the AˆRit from equation (10),

for every time a football match occurs. For every N football matches, the aggregated abnormal index returns for time period t is:

∑

= = N i it it AR N AR 1 ˆ 1 ( 12 )

In which the variance is:

∑

= = N i it _N i AR 1 2 2 1 ) var( σ_ε ( 13 )

In which ARit denotes the average abnormal return. Using these estimates, the average abnormal

return for the index can be used to see the effect of the international football matches the trading day t after the event during a World Cup or a European Championship.

(16)

∑

= = N i it it VAR AR AAR SD 1 ) ( ) ( ( 14 )

Subsequently, because of the relevant sample size, a students t-test is performed. To test the

alternative hypothesis, the t-test is used. TheH0is tested by using.

) ( _it it AAR SD AR = θ ( 15 ) 3.4 Robustness Checks

Maybe there are more relationships in the subsamples of this dataset. The relationship between

the top seven nations11_{in the world and the stock returns on the nations stock exchange will be}

tested analogously. Also, the effect on the stock market returns at country’s leading stock exchange of European teams at the World Cup will be tested.

Furthermore, the hypothesized relationship, as discussed in the literature review, can be time dependant or may decline overtime. Therefore, it is useful to see if there is an effect in time, more specifically between 1974-1996 and 1998 and 2008. This time period is chosen to split the dataset up in two most even parts.

Outliers are defined by a “large” deviation in the residuals. Anderson et al. (2002) suggest that an outlier with a z-score bigger than -3 and +3 should be defined as an outlier. In other words, if the trading day after a win or loss exhibits large negative or positive residual returns, then this observations can be excluded from the sample. This is approximately the upper 1% and the lower 1% of the sample.

(17)

IV. Data

The sample contains the wins and losses of football matches played during the World Cup and the European Championship. The results of the World Cup of 1974 till the European

Championship of 2008 are collected from Wikipedia12_{. Also the number of goals scored are}

obtained from Wikipedia. The stock quotations are represented by the daily closing prices of the country’s leading stock exchange, obtained from DataStream, if available from 1974. Representing the returns of the country’s leading stock exchange of the countries that played during the World Cup and European Championship. The list per country with wins and losses per country and mean returns can be found in Appendix A.

The data is not vulnerable towards data snooping. This is because the probability that one team will win and the other team will lose is very substantial. The timeframe for this thesis is the

period January 1st_{1973 till October 16}th_{2008. The condition to be included in the sample is that}

the national football team has a stock exchange in their country and should win or loose a football match during the World Cup or European Championship.

Table 1: Descriptive Statistics.

The table reports the wins and losses of international football matches played at the World Cup or at the European Championship. The football matches are played between 1974 and 2008, dependant on data availability. Every four

year a World Cup or European Championship is played. World Cup and European Championship will follow each other up after two years. The mean return reported in this table is the normal return at the country’s leading stock exchange the trading day after the international football match. The wins and losses per country are available in

(18)

All Games World Cup European Championship Wins Observations 321 209 112 Mean 0.03% 0.05% -0.005% Min -3.91% -3.91% -2.95% Max 6.83% 6.83% 3.02% Std. Devation 1.21% 1.34% 0.90% Kurtosis 7.24 6.73 4.43 Jarque Bera 256 134 10 Losses Observations 237 134 103 Mean -0.13% -0.27% 0.05% Min -6.90% -6.90% -2.95% Max 8.25% 2.40% 8.25% Std. Devation 1.35% 1.29% 1.40% Kurtosis 16.40 8.36 21.38 Jarque Bera 1853 217 1674

A total of 558 football matches on a World Cup and European Championship are classified, 321 winning matches and 237 losing matches, as relevant (mood) events. The losses have a mean return -0,13% and the wins have a mean return of 0,03%.

V. Results

First the result of the regression model will be presented and interpretated. After that the results will be compared with the event study model. At last, the results will be checked for robustness.

5.1 Regression model

(19)

Table 2: Abnormal stock returns the trading day after an international football match.

This table reports the Ordinary Least Squares estimates of the βW and βL fromεˆit =βWWit +βLLit +υit

defined by the regression of Edmans et al. (2007). The t-statistic shows the statistical significance of the coefficients.

All Games World Cup European Championship Wins Observations 321 209 112 Coefficient 0.0008 0.0010 0.0003 T-statistic 1.18 1.28 0.25 Losses Observations 237 134 103 Coefficient -0.0010 -0.0022 0.0007 T-statistic -1.26 -2.19 ** 0.58

***, ** and * Indicate statistical significance at the 1, 5 and 10%-levels, respectively.

All Games World Cup and European Championship

R-Squared 1.20% 1.15%

Adj. R-Squared 0.69% 0.97%

Durbin Watson 1.866 1.83

White Heteroskedasticity Test

Observerd R-Squared 0.80 3.72

Prob. Chi-Squared 0.37 0.29

(20)

all games and for the World Cup is 0,08% and 0,10% respectively. However, these estimates are not significant distinguishable from zero. In other words, there is not a significant win effect for international football matches at the country’s leading stock exchange. The larger significant loss effects and the smaller insignificant win effects is consistent with the asymmetric relationship, according to the prospect theory (Kahneman and Tversky, 1979), discussed in the literature review, between wins and losses.

The R-Squared and the Adjusted R-Squared of the OLS are very low with 1.2% and 0,7% for all games and for the World Cup and the European Championship respectively. So the model badly explains the relationship between the dependant and independent variables. The same applies for R-Squared and the Adjusted R-Squared for the World Cup and European Championship. The residuals are tested for autocorrelation. The value of the Durbin Watson statistic is 1,866 for all games and 1,83 for the World Cup and the European Championship. This gives no immediate indication of autocorrelation. Therefore, the null hypothesis of no autocorrelation cannot be rejected. To test whether the residuals are heteroskedastic the White’s heteroskedasticity test is used. The White’s test gives a probability of the Chi-Square of 0,37 for all games and 0,29 for the World Cup and the European Championship. This gives no indication of heteroskedasticty and therefore the null hypothesis of homoskedasticity cannot be rejected.

(21)

0,25% at a 5% level and the OLS Coefficient on the loss dummy for the World Cup is -0,21% at a 10% level. The OLS estimation gives no significant coefficients for the loss or win dummies for the European Championship.

Table 3: Abnormal stock returns the trading day after an international football match based on difference in goals scored or conceived.

defined by the regression of Edmans et al. (2007). This table is about the difference in goals scored. This is seperated into two groups: a “small difference” which is a difference of 0 or 1 goal(s) and a “big difference”

which is a difference of two or more goals. The t-statistic shows the statistical significance of the coefficients.

All Games World Cup European Championship Small Difference Big Difference Small Difference Big Difference Small Difference Big Difference

Wins Observations 183 138 114 95 45 43 Coefficient 0.0020 -0.0009 0.0025 -0.0007 0.0017 0.0012 T-statistic 2.34 ** -0.88 2.28 ** -0.60 0.93 0.87 Losses Observations 151 86 87 47 64 39 Coefficient -0.0012 -0.0006 -0.0021 -0.0025 0.0000 -0.0012 T-statistic -1.24 -0.46 -1.65 * -1.46 0.02 -0.70 ***, ** and * Indicate statistical significance at the 1, 5 and 10%-levels, respectively.

All Games World Cup and

European Championship

R-Squared 1.42% 2.23%

Observerd R-Squared 8.13 ** 16.49 **

ARCH LM ( 10 lags)

(22)

Table 3 shows a significant relationship between a small difference in goals scored or conceived for international football matches played on the World Cup and returns at the country’s leading stock exchange. An explanation behind this might be that a the small difference in goals scored or conceived, will cause a more uncertain end result of the match. Tension is becoming bigger and as a result, emotions may be more stronger after such matches. Furthermore, the asymmetric relationship found in Table 2 has been changed and has become more symmetric in table 3.

The Durbin Watson statistic shows again a value of 1,83 for all games and 1,85 for the World Cup and the European Championship. Which is no indication of autocorrelation The White’s test indicates that there might be heteroskedasticity, however, after testing for ARCH effects, there are no significant ARCH effects found. So a GARCH (1,1) test or EGARCH test will not be performed.

The results in Table 3 are concerned about the difference in goals scored or conceived. Table 4 presents the impact of the amount of goals scored. A small amount of goals scored or conceived, being 0 or 1, or big amount of goals scored or conceived, being 2 or more. The OLS coefficient on the dummy for all games which are lost by a small amount is -0,29% at a 5% level. If we split all the games up into World Cup and European Championship the OLS Coefficient for the World Cups’ on the loss dummy is even more strong with a -0,47% on a 5% level. While, the OLS Coefficient for the European Championship on the loss dummy is -0,06%, which is insignificant. The OLS coefficients for the win dummies are almost indistinguishable from zero and also insignificant.

(23)

Table 4: Abnormal stock returns the trading day after an international football match based on the amount of goals scored or conceived.

defined by the regression of Edmans et al. (2007). This table reports about the amount of gaols scored. This is seperated into two groups: a “small amount” which is a difference of 0 or 1 goal(s) and a “big amount” which

are two or more goals. The t-statistic shows the statistical significance of the coefficients.

All Games World Cup European Championship Small Amount Big Amount Small Amount Big Amount Small Amount Big Amount

Wins Observations 285 36 190 19 95 17 Coefficient 0.0012 0.0006 0.0014 0.0009 0.0007 0.0002 T-statistic 0.92 0.82 0.92 0.93 0.29 0.12 Losses Observations 73 164 41 93 32 71 Coefficient -0.0029 -0.0001 -0.0047 -0.0012 -0.0006 0.0013 T-statistic -2.09 ** -0.12 -2.54 ** -0.95 -0.29 0.89 ***, ** and * Indicate statistical significance at the 1, 5 and 10%-levels, respectively.

R-Squared 1.05% 1.79%

(24)

return for such a conditional loss, the trading day after a match. This can be the case because a small difference or a small amount of goals scored or conceived in an international football match brings more uncertainty of the expected outcome. Therefore, the tension might be much bigger when the game has “a close call” and the emotions released after the match can be much higher.

5.2 Event Study

In this paragraph the results from the regression model from Table 2 can be compared with the event study model. First, the effect is computed between an international football match on the World Cup and the European Championship and the returns at the country’s leading stock exchange the trading day after the match. Table 5 shows the abnormal returns for both the wins and losses.

Table 5: Abnormal stock returns the trading day after an international football match

This table reports the abnormal returns using the event study model used by Brown & Warner (1985) and MacKinlay (1997). Whereby the abnormal returns are calculated.

All Games World Cup European Championship Wins Observations 321 209 112 Abnormal Return 0.047% 0.048% 0.045% T-statistic 0.81 0.60 0.63 Losses Observations 237 134 103 Abnormal Return -0.168% -0.345% 0.062% T-statistic -2.05 ** -3.35 *** 0.48

(25)

games, for losses at international football matches is -0,17% being significant at a 5% level. For international football matches played at the World Cup the abnormal return is -0,34%.

Both models conclude that there is not a significant win effect for international football matches at the country’s leading stock exchange. However, the loss effect at the World Cup is economically twice as strong and also statistically significant. The event study model generates more economically and statistically more significant results than the regressions form the Edmans et al. (2007) model.

5.3 Robustness Checks

In Appendix B the results of the robustness check for the international matches played at the World Cup and European Championship can be found. The results do not change. By removing the outliers the same results are found. Furthermore, the OLS coefficient on the win dummy is still insignificant.

5.3.1 Time Dependant

(26)

Table 6: Abnormal stock returns the trading day after an international football match based on different time periods

This table reports the OLS coefficients on the win and the loss dummies from εˆit =βWWit +βLLit +υit defined

by the regression of Edmans et al. (2007). The data is divided in to two groups. The first group from 1974 to 1996 and the second group from 1998 tot 2008. The t-statistic shows the statistical significance of the coefficients.

All Games World Cup European Championship 1974-1996 1998-2008 1974-1996 1998-2008 1974-1996 1998-2008 Wins Observations 155 166 103 106 52 60 Coefficient 0.0014 0.0002 0.0012 0.0009 0.0016 -0.0009 T-statistic 1.43 0.26 1.06 0.76 0.99 -0.57 Losses Observations 111 126 72 62 39 64 Coefficient -0.0006 -0.0013 -0.0020 -0.0025 0.0020 -0.0002 T-statistic -0.54 -1.23 -1.46 -1.64 1.08 -0.11

R-Squared 0.17% 1.61%

(27)

5.3.2 Top Seven football nations

(28)

Table 7, Abnormal stock returns the trading day after an international football match based on the Top Seven football nations.

defined by the regression of Edmans et al. (2007). The top seven football nations consist out of five European Nations and two South American nations: Germany, Italy, France, England, Spain, Argentina and Brazil, which is

Panel A. Panel B includes the other 20 (football) nations.

All Games World Cup European Championship

Wins

Panel A: Top Seven Football Nations

Observations 198 137 61 Coefficient 0.0014 0.0020 -0.0002 T-statistic 1.63 2.03 ** -0.10 Panel B: Other Football Nations

Observations 123 72 51 Coefficient -0.0002 -0.0009 0.0008 T-statistic -0.15 -0.61 0.49 Losses

Observations 87 50 37 Coefficient -0.0022 -0.0029 -0.0011 T-statistic -1.71 * -1.76 * -0.59 Panel B: Other Football Nations

Observations 150 84 66 Coefficient -0.0003 -0.0018 0.0017 T-statistic -0.28 -1.41 1.17

(29)

R-Squared 1.02% 2.06%

The results in Table 7 indicate that the asymmetric relationship between wins and losses on the World Cup is becoming less skewed. Again there are no significant results found on the OLS coefficients for the loss and win dummies for international football matches played at the European Championship. This gives thoughts that one might hypothesize that there is a relationship between non European nations, playing at the World Cup, and the returns on the country’s leading stock exchange. This is because only European nations play on the European Championship. Therefore, it is useful to extend the findings of Table 7.

5.3.3 European teams on the World Cup

(30)

Table 8: Abnormal stock returns the trading day after an international football match based of the matches played for the National Teams from the European Union on a World Cup.

defined by the regression of Edmans et al. (2007). The European Nations and the other nations in the world are seperated in this table.

All Games European Nations Other Nations

Wins Observations 209 171 38 Coefficient 0.0010 0.0007 0.0027 T-statistic 1.28 0.76 1.39 Losses Observations 134 103 31 Coefficient -0.0022 -0.0022 -0.0023 T-statistic -2.20 ** -1.90 * -1.11

All Games European Nations and the Other Nations

R-Squared 1.15% 1.31%

Prob. Chi-Squared 0.73 0.015 **

ARCH Heteroskedasticity Test

Observerd R-Squared x 0.116

Prob. Chi-Squared x 0.733

(31)

Table 8. This means that for European nations there is a relationship between international football matches played at the World Cup and the return at the country’s leading stock exchange the trading day after the match. By removing the outliers, as can be found in Appendix G, this statement is confirmed at a 5% level. While there is an indication of heteroskedasticity using the White’s heteroskedasticity test, the ARCH test does not confirm that there are ARCH effects. So, no GARCH (1,1) or EGARCH has to be used.

5.3.4 Robustness check for the difference in goals scored and the amount of goals scored

After removing the outliers from equation (4) and equation (5) the results in table 7 become even more robust. For the amount of goals scored corrected for outliers, as can be found in Appendix D, the OLS Coefficient on the loss dummy is -0,55% and significant at a 1% level, this is the strongest (loss) effect found in this thesis.

5.3.5 The event study

By removing the upper 1% and the under 1% of the outliers, the abnormal returns become more economically and statically more significant, as can be seen in Appendix H. For all games and the matches at the World Cup the abnormal returns are now significant at a 1% level. The European Championship is significant at a 5% level. Without the outliers the World Cup has the strongest economically abnormal returns with a -0.215%.

VI Discussion

(32)

There is one exception for which we cannot accept the alternative hypothesis that a losing match has a bigger impact than a winning match on the returns of the country’s leading stock exchange the trading day after the match. Because, the difference in goals scored at the World Cup gives an abnormal return of 0,25% for a winning match and an abnormal return of -0,21% for a losing match. All the other test results from this thesis accept the alternative hypothesises that a losing match have a bigger impact than a winning match on the stock return of the country’s leading stock exchange. Therefore, we can reject the null hypothesis, that a losing match has the same impact as a winning match.

A football result at a World Cup in which 0 or 1 goal(s) are scored or a difference in goals of 0 or 1 goal(s) is encountered, then there is also a significant loss effect measured at the country’s leading stock exchange the trading day after the match. The loss effect at a World Cup whereby a small amount of goals are scored generate an abnormal return of -0,12%. The loss effect at a World Cup, whereby the difference in goals is a small amount, generates an abnormal return of -0,21%. The alternative hypothesis that the amount of goals scored or the difference in goals scored, in both winning and losing matches, will have an impact on the returns of the country’s leading stock exchange the trading day after the match cannot be rejected.

The approach used by Brown and Warner (1985) and MacKinlay (1997) explains whether wins and losses result in variation in stock performance. This gives a significant loss effect of -0,17%) and, when only the World Cup matches are taken into account, a highly significant loss effect of -0,34% is found.

(33)

The major pattern in the observations is that all the observations are in one month every two years from 1974 till 2008. The most common relationships are the loss effects found at the World Cups. Whereas, there are no effects found at the European Championships. Furthermore, a relationship is found in football matches at the World Cups where there is a small amount of goals conceived or where there is small difference in goals scored or conceived. There are two win effects found in this thesis. First, at the World Cup in the subsample of the top seven football nations in the world. Second, the win effect found at the World Cup concerning the small difference in goals scored.

The mechanisms underlying these patterns are that people tend to misattribute their mood to different or wrong features of the environment. This thesis show that, when the national football loses it causes a bad mood by investors and as a result more negative in evaluating propositions. The hypothesis is acknowledged in the sense that this misattribution stretches out to investments and subsequently stock prices.

The results are show the same results as the research performed by Edmans et al. (2007). There is indeed a strong loss effect (Edmans et al., 2007; Scholtens and Peenstra, 2008; Renneboog and Vanbrabant, 2000) and Edmans et al. (2007) also find insignificant win effects in their investigation as opposed to Scholtens and Peenstra (2008). However, as oppossed to the Edmans et al. (2007) research, in the subset “top seven football nation” Edmans et al. (2007) found no win effects. This thesis, however, does find a win effect for the World Cups. Furthermore, Zuber (2005) does not find any effects concerning the difference in goals scored or conceived for matches played by English public traded football clubs in the Premier League. This thesis does find win and loss effects.

(34)

significant results both in terms of wins and losses. This was not acknowledged in previous research.

VI. Conclusion

Psychological literature predicts that the result of international football matches causes an up- or down beat in mood. This thesis examines the relationship between the outcome of an international football match and the stock market reaction the trading day after the match at the country’s leading stock exchange, from 1974 to 2008. Whereby the outcome of international football matches that are played at the World Cup and the European Championship are used as a proxy for the mood variable.

The most important finding is that a significant loss effect is encountered at a World Cup of -0,22%. When a game at the World Cup is lost with a small difference in goals, when 0 goals or 1 goal are conceived, the strongest effect is encountered. This gives an economically and statistically significant loss effect of -0,47%. This thesis interpretates that the loss effect is a result from international football matches and stretches out to the mood of the investor and subsequently the stock prices. This acknowledged the belief that market participants can earn systematically abnormal returns on their investment.

(35)

errors would be more downward biased. 4) Datastream has limited availability towards index data of non European countries before 1990. Therefore, the data before 1990 is underweighted in the dataset. 5) This thesis use a full day of trading as the measurement day when the football result is known. However, some stock exchanges are open when a football match is played during the week. So, when a football match is played during the week and the nations stock exchange is open, an effect can already be incorporated in the prices. This can cause errors in the accuracy of the results found in this thesis.

(36)

References

• Anderson, D. R., Sweeney, D. J., and Williams, T. A., 2002. Statistics for Business and

Economic, 8th edition, West Publishing Company.

• Ashton, J., Gerrard, B., and Hudson, R., 2003, Economic impact of national sporting

success: Evidence from the London Stock Exchange, Applied Economic Letters 10, 783– 785.

• Avramova, Y. R., Stapel, D. A., Moods as Spotlights: The Influence of Mood on Accessibility Effects, Journal of Personality & Social Psychology, Sep2008, Vol. 95 Issue 3, p542-554

• Berthier, F., and Boulay, F., 2003, Lower myocardial infarction mortality in French men

the day France won the 1998 World Cup of football, Heart 89, 555–556

• Boyle, G., and Walter, B. 2002, Reflected glory and failure: International sporting success

and the stock market, Applied Financial Economics 13, 225–235.

• Brown, S.J. and Warner, J.B., 1985. Event Studies with Daily Returns, Journal of

Financial Economics, 14(1), 3-31

• Cao, M. and Wei, J., 2005, Stock market returns: A note on temperature anomaly,

Journal of Banking and Finance, 29, 1559–1573.

• Carroll, D., Ebrahim, S., Tilling, K., Macleod, J. and Smith, G.D., 2002, Admissions for

myocardial infarction and World Cup football: Database survey, British Medical Journal 325, 1439–1442.

• Chi, J., and Kloner, R., 2003, Stress and myocardial infarction, Heart, 89, 475–476

• Edmans, A., García, D., Norli, Ø., 2007, Sports Sentiment and Stock Returns, Journal of

Finance, Vol. 62 Issue 4, p1967-1998, 32p

• Fama, E. F., 1970, Efficient Capital Markets: A Review of Theory and Empirical Work,

Journal of Finance, Vol. 25 Issue 2, p383-417, 35p

• Forrest, D., Goddard, J. A. and Simmons, R., 2005, Oddsetters as forecasters: the case of

English football, International Journal of Forecasting, 21, 551–64.

• French, K., Poterba, J., 1991, “Investor Diversification and International Equity Markets”,

(37)

• Glascock, J., Henderson, G., Shah, V. and Officer, D., 1991, Sensitivity of the SCPE test statistic, Journal of Economics and Business, 43, 49–57.

• Hirt, E. R., Erickson, G.A. Kennedy, C., and Zillmann, D., 1992, Costs and benefits of

allegiance: Changes in fans’ self-ascribed competencies after team victory versus defeat,

Journal of Personality and Social Psychology 63, 724–738.

• Hirshleifer, D., 2001, Investor psychology and asset pricing, Journal of Finance 91, 342–

346

• Hirshleifer, D., and Shumway, T. 2003, Good day sunshine: Stock returns and the weather, Journal of Finance 58, 1009–1032.

• Huberman, G., 2001, Familiarity Breeds Investment, Review of Financial Studies, 2001,

Vol. 14 Issue 3, p659, 22p

• Isen, A. M. 1984, Toward understanding the role of affect in cognition. In R. S. Wyer &

T. K. Srull (Eds.), Handbook of social cognition, Vol. 3, pp. 179-236. Hillsdale, NJ: Erlbaum.

• Isen, A., Shalker, T., Clark, M., and Karp, L., 1978, “Affect, Accessibility of Material in

Memory and Behavior: A Cognitive Loop?”, Journal of Personality and Social

Psychology, 36, 1-12.

• Kahneman, D., Tversky, A. 1979, “Prospect Theory: An Analysis of Decision Under

Risk.” Econometrica, 47, 263-291.

• Markman, K. and Hirt, E., 2002, Social prediction and the “allegiance bias,” Social

Cognition, 20, 58–86.

• Mackie, D.M., and Worth, L.T., Forgas, J., 1991, Feeling Good, but not thinking straight:

The impact of positive mood on persuasion, Emotions and Social Judgements, Oxford

• Nofsinger, J.R., 2005, Social Mood and Financial Economics, Journal of Behavioral

Finance, Vol. 6 Issue 3, p144-160.

• Nofsinger, J.R., 2007. The Psychology of Investing, 3th edition. New Jersey: Pearson

Education

• Palomino, F., Renneboog, L. and Zhang, C., 2005, Stock price reactions to short-lived

public information: the case of betting odds, ECGI, Working Paper No. 81.

(38)

• Renneboog, L. and Vanbrabant, P., 2000, Share price reactions to sporty performance of soccer clubs listed on the London Stock Exchange and the AIM, Economic Research

Working Paper, No. 19.

• Ross, S.A., Westerfield, R.W. and Jaffe, J., 2005. Corporate Finance, international edition. Boston: McGraw – Hill

• Saunders, Edward M., 1993, Stock prices and Wall Street weather, American Economic

Review, 83, 1337–1345.

• Scholtens, B. and Peenstra, W., 2008, 'Scoring on the stock exchange? The effect of

football matches on stock market returns: an event study', Applied Economics, 99999:1

• Schwarz, N., & Clore, G.L., 1983. Mood, misattribution, and judgement of well-being:

Informative and directive functions of affective states. Journal of Personality and Social

Psychology, 45, 513–523.

• Stadtmann, G., 2006, Frequent news and pure signals: evidence of a publicly trade football club, Scottish, Journal of Political Economy, 53, 485–504.

• Wann, Daniel, Merrill Melnick, Gordon Russell, and Dale Pease, 2001, Sport Fans: The

Psychology and Social Impact of Spectators, (Routledge London).

• White, H., 1980, A Heteroskedasticity Consistent Covariance Matrix Estimator and a

Direct Test for Heteroskedasticity, Econometrica, Vol. 48, pp. 817-38.

• Wright, F., Bower, H., 1992, “Mood Effects on Subjective Probability Assessment”,

Organizational Behavior and Human Decision Process, 52, 276-291.

• Wyer, R.S., & Srull, T.K., 1986, Human cognition in its social context, Psychological

Review, 93, 322-359.

• Yuan, Kathy, Lu Zheng, and Qiaoqiao Zhu, 2006, Are investors moonstruck? Lunar

phases and stock returns, Journal of Empirical Finance 13, 1–23.

• Zuber, R. A., Yiu, P., Lamb, R. P. and Gandar, J. M., 2005, Investor-fans? An examination of the performance of publicly traded English Premier League teams,

(39)

Appendix A:

The mean of the stock returns at the country’s leading stock exchange the trading day after an international football match is played.

the number of wins and losses in international football matches.

Country Begin Time Serie Mean Return Mean Abnormal

Return Wins Losses

(40)

Appendix B:

Robustness check for the abnormal stock returns at the country’s leading stock exchange the trading day after an international football match.

defined by the regression of Edmans et al. (2007). The t-statistic shows the statistical significance of the coefficients. The upper 1% of the residuals and the lower 1% of the residuals are removed, called the outliers ,from the sample.

While there is an indication of heteroskedasticity using the White’s test, there are no ARCH effects found in this subdataset. Therefore the OLS estimates are used.

All Games World Cup European Championship Wins Observations 321 209 112 Coefficient 0.0006 0.0008 0.0003 T-statistic 0.92 0.96 0.26 Losses Observations 237 134 103 Coefficient -0.0013 -0.0026 0.0004 T-statistic -1.69 * -2.53 ** 0.31

R-Squared 2.41% 3.07%

Observerd R-Squared 412.10 *** 410.43 ***

(41)

Appendix C:

Robustness check for the abnormal stock returns at the country’s leading stock exchange the trading day after an international football match for the difference in goals scored or

conceived.

This table reports an extended version of the Ordinary Least Squares estimates of the βW and βL from it it L it W it β W β L υ

εˆ = + + defined by the regression of Edmans et al. (2007). The upper 1% of the residuals and the lower 1% of the residuals are removed, called the outliers, from the sample. While there is an indication of heteroskedasticity using the White’s test, there are no ARCH effects found in this subdataset. Therefore the OLS

estimates are used.

All Games World Cup European Championship Small Difference Big Difference Small Difference Big Difference Small Difference Big Difference

Wins Observations 183 138 114 95 45 43 Coefficient 0.0018 -0.0010 0.0021 -0.0008 0.0012 -0.0012 T-statistic 2.07 ** -0.97 1.95 * -0.70 0.88 -0.70 Losses Observations 151 86 87 47 64 39 Coefficient -0.0015 -0.0010 -0.0024 -0.0027 -0.0001 0.0012 T-statistic -1.55 -0.76 -1.95 * -1.61 -0.10 0.64 ***, ** and * Indicate statistical significance at the 1, 5 and 10%-levels, respectively.

R-Squared 3.20% 3.93%

ARCH LM ( 10 lags)

(42)

Appendix D:

Robustness check for the abnormal stock returns at the country’s leading stock exchange the trading day after an international football match for the amount of goals scored or

conceived.

εˆ = + + defined by the regression of Edmans et al. (2007). The upper 1% of the residuals and the lower 1% of the residuals are removed, called the outliers, from the sample. While there is an indication of heteroskedasticity using the White’s test, there are no ARCH effects found in this subdataset. Therefore the OLS

estimates are used.

Small Amount Big Amount Small Amount Big Amount Small Amount Big Amount

Wins Observations 285 36 190 19 95 17 Coefficient 0.0010 0.0004 0.0012 0.0006 0.0007 0.0002 T-statistic 0.83 0.57 0.80 0.61 0.30 0.12 Losses Observations 73 164 41 93 32 71 Coefficient -0.0034 -0.0004 -0.0055 -0.0013 -0.0006 0.0008 T-statistic -2.44 ** -1.78 -3.00 *** -1.06 -0.29 0.54

R-Squared 3.01% 1.79%

(43)

Appendix E:

Robustness check for the abnormal stock returns at the country’s leading stock exchange the trading day after an international football match for the time period.

εˆ = + + defined by the regression of Edmans et al. (2007). The upper 1% of the residuals and the lower 1% of the residuals are removed, called the outliers, from the sample. While there is an indication of heteroskedasticity using the White’s test, there are no ARCH effects found in this subdataset. Therefore the OLS estimates are used. The data is devided in to two groups. The first group from 1974 to 1996 and the second group

from 1998 tot 2008.

All Games World Cup European Championship 1974-1996 1998-2008 1974-1996 1998-2008 1974-1996 1998-2008 Wins Observations 155 166 103 106 52 60 Coefficient 0.0013 0.0000 0.0011 0.0004 0.0016 -0.0009 T-statistic 1.37 -0.04 0.97 0.39 1.00 -0.58 Losses Observations 111 126 72 62 39 64 Coefficient -0.0009 -0.0016 -0.0023 -0.0028 0.9368 -0.0005 T-statistic -0.81 -1.57 -1.70 * -1.89 * 1.08 -0.35

R-Squared 2.63% 3.49%

(44)

Appendix F:

Robustness check for the abnormal stock returns at the country’s leading stock exchange the trading day after an international football match for the Top Seven football nations.

εˆ = + + defined by the regression of Edmans et al. (2007). The upper 1% of the residuals and the lower 1% of the residuals are removed, called the outliers, from the sample.

Wins

Observations 198 137 61 Coefficient 0.0013 0.0020 -0.0002 T-statistic 1.57 1.96 * -0.11

Panel B: Other Football Nations

Losses

Observations 87 50 37 Coefficient -0.0026 -0.0036 -0.0011 T-statistic -2.04 ** -2.19 ** -0.59

Panel B: Other Football Nations

R-Squared 3.19% 4.10% Adj. R-Squared 2.49% 2.70% Durbin Watson 1.82 1.83

Observerd R-Squared 415.00 *** 431.1449 *** Prob. Chi-Squared 0.00 0.00

Observerd R-Squared 1.15 1.42 Prob. Chi-Squared 0.28 0.23