The influence of international soccer
sentiment on stock returns
A European perspective
Master Thesis MSc Finance
Abstract
This papers investigates the relationship between international soccer results and stock market returns. Other studies have already shown that there is a relationship between (international) soccer game losses and negative stock (market) returns or (international) soccer game wins and positive stock (market) returns. This research investigates and gives an answer to the question whether or not this relationship exists for Europe, based on a data sample containing price index returns and international soccer game results of 10 European countries from January 1983 until September 2014. The main finding of this research is that there is no evidence for a relationship between international soccer results and the market index returns of the 10 European countries.
2
1. Introduction
Traditional finance models have a basis in economics in which neoclassical economics is the dominant concept. In this concept, firms and individuals are agents who try to make optimal decisions with respect to constraints on resources and the price of assets is determined on a market which is subject to demand and supply. There are some fundamental assumptions about agents in neoclassical economics: (Ackert and Deaves, 2010)
Agents have rational behaviour in decision making.
Individuals maximize utility and firms maximize profits.
Agents act independently on the basis of relevant information.
However, recent literature has challenged these assumptions. Stracca (2004) reviews existing literature and finds several anomalies which result from behavioural biases in numerous studies. While anomalies are discovered, there is still no conclusive evidence that neoclassical economics is a flawed framework for studying the behaviour of agents. However, behavioural finance is still a rapidly growing field that deals with these challenges and provides explanations for why people make irrational financial decisions. It uses insights from psychology to understand how human behaviour influences the decisions of individual and professional investors, markets, and managers.
Within this field of finance a lot of recent papers, like Kamstra et al. (2000), Hirshleifer and Shumway (2003), and Edmans et al. (2007), examine the relationship between investor mood (sentiment) and stock prices. In these studies there are mainly two approaches in examining this relationship: an event study approach or a regression analysis on the relation between sentiment and stock prices.
3
strategies, but to document that mood can have an effect on market returns within Europe. This goal is similar to the one of Edmans et al. (2007). In contrast to Edmans et al. (2007) this study uses a more recent dataset and the data will only be extracted for Europe, in order to see whether the effect is significant within Europe. In addition the possible differences in the impact of international soccer results on stock markets between the 10 European countries of the sample will be examined.
Based on the motivation described, the research question is:
Are the stock market returns of the leading European indices significantly affected by international soccer game results?
The sub questions for this research are:
1. Is the impact of international soccer results on the stock market returns significant for Europe as a whole?
2. Are the stock market returns, for each country separate, significantly affected by the international soccer results of that country?
3. Is the impact of international soccer results on stock market returns different for northern and southern European countries?
4. Is the impact of international soccer results on stock market returns significantly different between European countries?
The choice for international soccer results as a measure of mood can be justified by the fact that soccer is of national interest in many countries around the world. This can be concluded by looking at the TV viewing figures, media coverage, and merchandise sales. Some examples, according to FIFA: More than 3.2 billion people, or 46.4 percent of the world population, watched the 2010 World Cup live for one minute or more, the average official viewing figure for each match of the 2010 World Cup was 188.4 million, the cumulative number of TV viewers of the 2006 World Cup was estimated to be 26 billion, and the 64 matches played by 32 countries in the 2010 World Cup were broadcasted in every territory in the world. It is hard to imagine other regular events that produce such substantial and correlated mood swings in a large proportion of a country’s population.
4
returns. In order to estimate the impact of wins and losses on stock market returns, regressions will be performed, in which will be controlled for the Monday and any other relevant effects.
This paper continues with a discussion of the relevant literature regarding this topic, which will eventually lead to the hypotheses. The next section will contain the data. After the data description, the methodology section is presented. In the following section the results of this study will be given and analysed. This research paper will be concluded with an answer to the research and sub questions.
2. Literature review
In this section the prior literature will be discussed, which eventually will lead to the hypotheses for this research.
2.1 Prior literature
According to the theoretical research of Qawi (2010) there are two main accepted directions within the field of the study of finance. On the one side there is the traditional and widely accepted approach of a fully rational agent and his decision making based on available data, and on the other side there is the psychological or behavioural approach which brings the element of human behaviour into the decision making process. More specific, behavioural finance uses insights from psychology research to examine how human behaviour influences the decision of individual and professional investors, managers and markets. It introduces many aspects of human behaviour into traditional finance to improve our understanding about financial analyst and investors.
The history of stock markets is full of events that are influential enough to give them a name. Examples are: the Go-Go years of the late 1960s, the black Monday crash of October 1987, and the internet bubble of the late 1990s. Each of these events incorporates a high level of stock market price changes, which seem to deviate from neoclassical economics. Therefore, according to the theoretical research of Baker and Wurgler (2007), researchers in behavioural finance have tried to enlarge the standard finance model with an alternative model that is built on two basic assumptions. The key assumption is that investors are influenced by sentiment. “Investor sentiment, defined broadly, is a belief about future cash flows and investment risks that is not justified by the facts at hand” (Baker and Wurgler, 2007, p. 129). The second assumption is that betting against investors that are influenced by sentiment is costly and risky. Therefore arbitrageurs (rational investors) are not as aggressive in forcing prices to fundamental values as the standard model would suggest.
5
finance papers. Various articles use the event study approach. Kamstra et al. (2000) use this approach in order to assess whether or not sleep might have consequences in financial markets. The idea behind this research is that relatively minor sleep imbalances have been shown to cause errors in judgement, impatience, loss of attention, and less efficient information processing of people and thus stock market participants. They find that a change from or to the daylight saving time (summer time), which is caused by clock shifting, which happens twice a year, implies a one-day loss of $31 billion on the NYSE, AMEX, and NASDAQ indices. This is called the daylight saving effect. Frieder and Subrahmanyam (2004) also use the event study approach to examine whether holidays that are not observed nationally in the U.S., have a significant impact on the U.S. stock market. They find that the anticipatory return effects of the two festive occasions studied are strong (St. Patrick’s day and Rosh Hashanah). The other approach, a regression analysis on the relation between sentiment and asset prices is for example used in the studies by Edmans et al. (2007) and Hirshleifer and Shumway (2003) which examine the influence of sunny weather on the leading stock exchanges in 26 different countries. It is predicted that sunny weather is associated with an upbeat mood. They find that sunshine is highly significantly correlated with daily positive stock returns.
In comparing the two approaches, clearly identifying a sudden change in the mood of investors, which gives a large signal-to-noise ratio in returns, is the main advantage of the event study approach. In contrast the main disadvantage of the event study approach in comparison to the regression analysis, is that the number of observed signals tends to be low. This reduces statistical power (Edmans et al., 2007).
6
of a country’s population. These characteristics provide strong a priori motivation for using game outcomes to capture mood changes among investors.” (p. 1968)
To support that sport evens have an impact on mood, literature has also investigated the impact of sport events on health. Carrol et al. (2002) find that admissions for acute myocardial infarction in England increased by 25% on 30 June 1998 and the two following days. On 30 June 1998 England played Argentina in the quarter final of the 1998 World Cup. The increase in admissions suggests that myocardial infarction can be triggered by severe mood swings, such as watching your country lose an important match on a World Cup. In contrast Berthier and Boulay (2003) hypothesise that an exceptional positive sporting event could decrease the total number myocardial infarction mortalities. They find that mortality from myocardial infarction of men in France was significantly lower on the day that France won the 1998 World Cup, compared with the five days before and after the final.
There is a body of literature that studies the influence of soccer results on stock returns. Within this field of research there are different types of studies. Some studies examine the influence of results of soccer clubs on their own stock prices. This is done for several countries including: England, Germany, Italy, Portugal and Turkey.1 Other studies examine international soccer results and its effect on local
market returns. Edmans et al. (2007) is an example of this kind of research. Another study done by Kaplanski and Levy (2010) examines whether international soccer results have an impact on the U.S. Market. This is called the spill over effect. The main findings of each type of study will be mentioned below.
2.1.1 The influence of soccer results of clubs on their own stock prices
Palomino et al. (2009) find evidence of investor overreaction following soccer game outcomes. Stock prices are sensitive to game results for 16 English soccer clubs that are listed on the London Stock Exchange (LSE) for the period 1999-2002. The average cumulative abnormal returns over a three day period are highly statistically significant for both wins and losses. Bell et al. (2012), who also examine stock price reactions to game results of 19 English soccer clubs, find that stock prices are affected by game results. However, these effects are modest compared with the changes in club stock prices caused by other variables. The market index has the biggest effect on stock returns. In addition they find that the importance of the game, measured in rivalry, or the closeness of the match to the end of the season, also have very modest impacts on stock returns. There are also findings that there is no relationship between returns and soccer results. Zuber et al. (2005) find that no material differences are found between on-season and off-season performance, despite the dramatically higher level of
1 Examples are: Palomino et al. (2009) and Bell et al. (2009) for England, Duque and Ferreira (2005) for Portugal,
7
information available to investors during the on-season period. The other researches on this specific topic have been performed for Portugal, Italy, Germany, and Turkey. The studies by Duque and Ferreira (2005) on Portuguese soccer clubs, Boido and Fasano (2007) on Italian soccer clubs, Stadtmann (2006) on German soccer clubs, and Demir and Danis (2011) on Turkish soccer clubs all show a relationship exists between the sporting success of soccer clubs and their stock performance.
Besides the study per country there are also studies done for soccer clubs within Europe, see e.g. Benkraiem et al. (2009), and Scholtens and Peentra (2009). Benkraiem et al. (2009) find that losses in particular cause significant drops in market prices. Wins do not have a significant influence on the stock market. In addition Scholtens and Peenstra (2009) find that stock markets react negative to losses and positive to wins. The response to losses however is much stronger than it is to wins. Furthermore, they find that stock markets react stronger to European matches than to matches in national leagues.
2.1.2 Influence of international soccer results on local markets
Edmans et al. (2007) consider 39 nations to examine the influence of international soccer results on the stock market. They document a strongly significant negative impact of losses by national soccer teams on the stock market. There is no evidence of a corresponding reaction after wins by the national teams. In addition Edmans et al. (2007) find that the effect is most striking for games in the World Cup, in specific for elimination games during the World Cup. Because of the magnitude of the loss effect, which is rather big, investor may think that they can obtain large excess returns by trading on these mood events, by taking short positions in stocks of both countries. However, Edmans et al. (2007) think that the events do not occur with enough frequency to justify holding a portfolio that is fully dedicated to these kind of trades. Ashton et al. (2011) consider whether international soccer results and stock market returns are linked for the UK stock market and find that the national soccer matches have an effect on the returns of the London Stock Exchange.
2.1.3 Influence of international soccer results on the U.S. market
8
2.2 Influence of cultural differences on the impact of international soccer results on stock market returns
As can be observed from the sub questions from chapter 1, this study is also going to analyse the differences in the impact of international soccer results on stock market returns between parts and countries of Europe. This is based on the expectation that there are cultural differences between parts or countries of Europe and that these cultural differences impact the mood of the European countries after international soccer games differently. No literature could be found that investigated the cultural differences between northern and southern European countries, but there was literature that investigated these differences for other countries within Europe. The study by Kolman et al. (2002) investigates the cultural differences between four central European countries (Czech Republic, Hungary, Poland and Slovakia) based on the Hofstede dimensions. These dimensions are power distance, individualism-collectivism, uncertainty avoidance, masculinity-femininity, and long versus short term orientation. The findings of the study by Kolman et al. (2002) are that there are important differences between the four central European countries and that there are cultural differences between central European countries and Western Europe (represented in their study by the Netherlands). Previous literature has thus shown that cultural differences do exist between European countries that are not part of the dataset of this study.
2.3 Hypotheses
As can be read most literature suggests that national or international win or loss soccer game results affect the stock prices or the stock market returns. In advance of this study it is expected that stock market returns of the leading European indices are significantly affected by international soccer game results, unless this research finds enough evidence that this is not true.
9
3. Data
In this section the data sample will be discussed. Furthermore the sample criteria for selecting relevant international soccer games and the descriptive statistics will be reported.
3.1 Sample selection
The dataset of this research contains data about the price index and the international soccer game results of 10 different European countries from January 1983 until September 2014. In addition it contains the European market index for the same time span. This time span is comparable with the research of Edmans et al. (2007), which is also about 30 years, but this time span is more recent. The countries that are part of the sample are: Denmark, England, France, Germany, Italy, Netherlands, Norway, Portugal, Spain, and Sweden. These countries are chosen because they are the most influential soccer countries of Europe. All but the northern countries are in the top 10 of best European soccer countries, according to www.eloratings.net. Elo ratings measures the ability of countries. The sample had to contain both northern and southern European countries, in order to be able to give an answer to sub question 3. Therefore also three northern European countries, which are not part of the top 10 of best European soccer countries, are added to the data sample. Now the number of northern countries is equal to the number of southern countries (Italy, Portugal and Spain) and chosen northern countries are the best of northern Europe, according to www.eloratings.net.
For each country the market index returns from January 1983 until September 2014 are calculated from the price index of the leading index of the country. These price indices of each country are extracted from DataStream. The total return index is not used because for some countries of the sample only a few years of total return index are given. Therefore using the total return index would result in eliminating a substantial part of the data. The expectation is that this does not provide a bias for this study because the possible mood effect can also be seen on a price index. The dividends which are reinvested in the total return index are expected to not be affected by international soccer returns, whereas to daily price movements are. In table 1 the details on the indices used in this study can been found.
10
Table 1 Indices details
This table reports the beginning of the time series and the mean return of the indices used in the countries that are part of the sample. In addition the number of win and loss games in each country are reported.
Country Index Time series begin Mean return Wins Losses
Denmark OMX Copenhagen 04/12/1989 0.029% 24 19
England FTSE 100 03/01/1983 0.025% 28 30
France CAC 40 09/07/1987 0.015% 32 18
Germany DAX 30 03/01/1983 0.035% 63 20
Italy Milan Comit 30 01/01/1993 0.012% 31 16
Netherland AEX 03/01/1983 0.028% 47 33
Norway OBX Oslo 02/01/1987 0.027% 15 18
Portugal PSI 20 01/01/1993 0.010% 33 22
Spain IBEX 35 05/01/1987 0.022% 43 19
Sweden OMX Stockholm 02/01/1986 0.034% 22 25
All countries 0.024% 338 220
The European market index is also extracted from DataStream. Because all the countries of this research are European, the European market index instead of the world market index is chosen to account for correlation across the countries of the sample.
3.2 Sample criteria for international soccer games
The sample criteria that are used during the search for relevant mood events can be seen in table 2.
Table 2 Sample criteria
This table shows the sample criteria used on the data of this research and the remaining number of games after the criterion.
Criteria Number of games
All games between 01-1983 - 09-2014 1880
Relevant games (Elo rating range criterion) 870
Price index return available for relevant games 755
Only Win and Loss games 558
After the first round of the data collection 1880 international soccer games were part of the sample. These were all games played by the 10 countries of the sample on World Cups, European Cups, and the qualifying stages for the World Cups and European cups.
11
predetermine. To identify “important and hard to predetermine” qualifying games the closeness in the ability of the two opponents, which is measured by Elo ratings, is used. These Elo ratings are collected from www.eloratings.net. The Elo rating system was originally developed for assessing the strength of chess players, but in time it has been widely adopted in various other sports, including soccer (Hvattum and Arntzen, 2010). A qualifying game is important, and thus added to the dataset, if the Elo rating of the two opponents is within 125 point (after adding 100 points to the team with the home advantage) or if it is a knock-out round between the qualifying stage and the group stage of the Word Cup or European Cup. This is the same method as used in Edmans et al. (2007). Due to this criterion 1010 games dropped from the sample.
The second criterion was that the price index return must be available for the day following the soccer game, in order to be able to calculate the return for that specific event. For some countries the price index return was only available from the mid 90’s. Therefore 115 games dropped from the sample. For games that are played on Friday, Saturday or Sunday, the return from the closing price on Friday to the closing price on Monday is used.
The last criterion was that only win and loss games were part of the sample, because the literature review did not show that draw games had significant influence on stock prices or stock market returns. This last criterion led to an additional reduction of 197 games from the sample.
3.3 Descriptive statistics
12
Table 3
Begin times of international soccer games
This table shows the number of games that start at or after 4.00 pm and the number of games that start before 4.00 pm for the countries Spain, Netherlands, Germany, and Sweden. The starting times where available from 1990 and are extracted from www.soccerway.com. In addition the percentage of games that start at or after 4.00 pm are given for each country.
Country Spain Netherlands Germany Sweden
Total Nr. of games 57 66 70 46
Nr. of games starting ≥ 4.00 pm 54 64 66 44
Nr. of games starting < 4.00 pm 3 2 4 2
% of games that start ≥ 4.00 pm 94.74% 96.97% 94.29% 95.65% From table 3 it can be seen that for each country of this investigation almost 95% of the games began at or after 4.00 pm. With this result it is decided that the returns will be calculated with the closing price of the index on the game day (t-1) and the closing price of the index on the day following the game day (t). It is believed that this will capture the eventual mood effect. These returns will be used during this study. The descriptive statistics about the returns that are explained above can be seen in table 4.
Table 4
Descriptive statistics of the returns after soccer games
The number of soccer game days and the mean returns, standard deviation of returns, the minimum return, and the maximum return after soccer games are reported. The mean returns, standard deviation, minimum return, and maximum return are given in percentages.
No games Game days
Wins Losses Number of days 70955 338 220 Mean returns 0.024% 0.045% -0.023% SD of returns 1.322% 1.256% 1.322% Minimum return -23.996% -4.087% -4.913% Maximum return 13.484% 5.722% 4.358%
As can be seen 70,955 trading days are not associated with a soccer game. The mean return and standard deviation for these trading days are 0.024% and 1.322%. Looking at the descriptive statistics after an international soccer win or loss, it can be seen that the average returns after a win are higher than after a loss. In addition table 4 shows that the minimum returns are lower after a loss game in comparison to a win game, and that the maximum returns are higher after a win game in comparison to a loss game.
4. Research method
In this section the research method for each sub question of this study will be reported.
13
expectations are not fulfilled, it can be interpreted that investors are rational, stock markets are efficient, and that the economic consequences associated with international soccer game results are too small to influence the stock market index of the country. In order to be able to estimate the impact of wins and losses on stock market indices, first the continuously compounded daily stock returns for each stock market index have to be calculated. The data for the continuously compounded daily stock returns are obtained from DataStream and the calculation for it, according to brooks (2008), is as follows:
𝑅𝑐𝑡 = 100% ∗ 𝐿𝑛 ( 𝑃𝑃𝑐𝑡
𝑐𝑡−1 ) (1)
In which 𝑅𝑐𝑡 is the continuously compounded daily return on the leading stock market index of country
c at trading day t, 𝑃𝑐𝑡 is the closing price of the price index for country c at trading day t, and 𝑃𝑐𝑡−1 is
the closing price of the price index for country c at trading day t-1.
The first sub question, which examines whether the impact of international soccer results on the market returns is significant for Europe as a whole, will be answered by performing a pooled ordinary least squares (OLS) regression with all countries of the sample. OLS regression analysis is chosen because regression analysis has high statistical power and is a widely applicable method. Due to the fact that the equation will be estimated simultaneously for all countries, the time subscript will drop from the regression because there are 10 different time spans (each country has their own specific time span). A solution is to treat each trading day of the countries as an event. So trading day 1 of country c, is now event 1 of country c. Therefore the time subscript (t) from the regression will be replaced by events which will be denoted by i. The regression that needs to be estimated is:
𝑅𝑐𝑖 = 𝛽0+ 𝛽1 𝑅𝑐𝑖−1+ 𝛽2 𝑅𝑚𝑖−1+ 𝛽3 𝑅𝑚𝑖+ 𝛽4 𝑅𝑚𝑖+1+ 𝛽5 𝑀𝑖+ 𝛽6 𝑊𝑐𝑖+ 𝛽7 𝐿𝑐𝑖+ 𝑢𝑖 (2) where 𝑅𝑐𝑖 is the continuously compounded daily return on the leading stock market index of country
c at event i, 𝛽0 is the constant, 𝑅𝑐𝑖−1 is the one event lagged index return, 𝑅𝑚𝑖 is the continuously
compounded return on the European market index at event i, 𝑀𝑖 is a dummy variable which equals
one for Monday and zero otherwise, 𝑊𝑐𝑖 is a dummy variable which equals one if country c wins a
soccer game for the return that is calculated from the closing price of the index on the game day(t-1) and the closing price of the index on the event following the game day (t) and zero otherwise, 𝐿𝑐𝑖 is a
dummy variable which equals one if country c loses a soccer game for the return that is calculated from the closing price of the index on the game day(t-1) and the closing price of the index on the event following the game day (t) and zero otherwise, and 𝑢𝑐𝑖 is the error term.
The one day lagged index return, 𝑅𝑐𝑖−1, is included to deal with first order serial correlation, which is
14
significant relationship between the daily stock market returns of the indices and their value the previous trading day. The continuously compounded return on the European market index is included for the correlation of local indices across the countries of the sample. While some local market indices may be lagging and others leading the European market index, 𝑅𝑚𝑖−1 and 𝑅𝑚𝑖+1 are included in the
regression. Due to the fact that many games are played between Friday afternoon and Sunday afternoon, the return on Mondays is a frequently measured observation. This may cause a false day of the week relationship between soccer game results and stock returns. Therefore the dummy variable 𝑀𝑖 is included to control for this Monday effect. The variables 𝑊𝑐𝑖 and 𝐿𝑐𝑖 are included to estimate the
effect of international soccer game results on European stock market indices.
After the regression of each country, and also for the other regressions to come, the outcomes will be checked on heteroscedasticity, using White’s general test for heteroscedasticity. If the residuals are found to be heteroscedastic, the regression equation will be estimated again with heteroscedasticity-robust standard errors. In addition all regressions will be tested on multicollinearity by looking at the matrix of correlations. If needed, one of the collinear variables will be dropped from the regression. In order to give an answer to sub question two, we estimate the impact of wins and losses on the stock market returns for each country (c) with the following OLS regression, which controls for the Monday and other possibly relevant effects:
𝑅𝑐𝑡 = 𝛽0𝑐+ 𝛽1𝑐 𝑅𝑐𝑡−1+ 𝛽2𝑐 𝑅𝑚𝑡−1+ 𝛽3𝑐 𝑅𝑚𝑡+ 𝛽4𝑐 𝑅𝑚𝑡+1+ 𝛽5𝑐 𝑀𝑡+ 𝛽6𝑐 𝑊𝑐𝑡+ 𝛽7𝑐 𝐿𝑐𝑡+ 𝑢𝑐𝑡
(3) Where 𝑅𝑐𝑡 is the continuously compounded daily return on the leading stock market index of country
c at trading day t, 𝛽0𝑐 is the constant, 𝑅𝑐𝑡−1 is the one day lagged index return, 𝑅𝑚𝑡 is the continuously
compounded return on the European market index on day t, 𝑀𝑡 is a dummy variable which equals one
for Monday and zero otherwise, 𝑊𝑐𝑡 is a dummy variable which equals one if country c wins an
international soccer game for the return that is calculated from the closing price of the index on the game day (t-1) and the closing price of the index on the day following the game day (t) and zero otherwise, 𝐿𝑐𝑡 is a dummy variable which equals one if country c loses an international soccer game
for the return that is calculated from the closing price of the index on the game day (t-1) and the closing price of the index on the day following the game day (t) and zero otherwise, and 𝑢𝑐𝑡 is the error term.
15
The last sub question will be answered by examining the differences, resulting from the impact of international soccer results on stock market returns, between European countries. This will be done by including dummy variables for each country in the regression equation (Newbolt et al., 2010). Due to the fact that all the countries are part of this regression the time subscripts will again be replaced by event subscripts. The following regression will be estimated:
𝑅𝑐𝑖 = 𝛽1 𝑅𝑐𝑖−1+ 𝛽2 𝑅𝑚𝑖−1+ 𝛽3 𝑅𝑚𝑖+ 𝛽4 𝑅𝑚𝑖+1+ 𝛽5 𝑀𝑖+ 𝛽6 𝑋𝑊𝑖,𝑑𝑒+ 𝛽7 𝑋𝑊𝑖,𝑒𝑛+ 𝛽8 𝑋𝑊𝑖,𝑓𝑟 + 𝛽9 𝑋𝑊𝑖,𝑔𝑒+ 𝛽10 𝑋𝑊𝑖,𝑖𝑡+ 𝛽11 𝑋𝑊𝑖,𝑛𝑒+ 𝛽12 𝑋𝑊𝑖,𝑛𝑜+ 𝛽13 𝑋𝑊𝑖,𝑝𝑜+ 𝛽14 𝑋𝑊𝑖,𝑠𝑝 + 𝛽15 𝑋𝑊𝑖,𝑠𝑤+ 𝛽16 𝑋𝐿𝑖,𝑑𝑒+ 𝛽17 𝑋𝐿𝑖,𝑒𝑛+ 𝛽18 𝑋𝐿𝑖,𝑓𝑟+ 𝛽19 𝑋𝐿𝑖,𝑔𝑒+ 𝛽20 𝑋𝐿𝑖,𝑖𝑡 + 𝛽21 𝑋𝐿𝑖,𝑛𝑒+ 𝛽22 𝑋𝐿𝑖,𝑛𝑜+ 𝛽23 𝑋𝐿𝑖,𝑝𝑜+ 𝛽24 𝑋𝐿𝑖,𝑠𝑝+ 𝛽25 𝑋𝐿𝑖,𝑠𝑤+ 𝑢𝑐𝑖
(4) where 𝑅𝑐𝑖 is the continuously compounded daily return on the leading stock market index of country
c at event i, 𝛽0 is the constant, 𝑅𝑐𝑖−1 is the one day lagged index return, 𝑅𝑚𝑖 is the continuously
compounded return on the European market index at event i, 𝑀𝑖 is a dummy variable which equals
one for Monday and zero otherwise, , and 𝑢𝑐𝑖 is the error term. The dummy variables, on the right
hand side of equation 4, for a win for the specific countries are specified as:
𝑋𝑊𝑖,𝑑𝑒= {1 if Denmark wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑒𝑛= {1 if England wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑓𝑟 = {1 if France wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑔𝑒= {1 if Germany wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑖𝑡= {1 if Italy wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑛𝑒= {1 if Netherlands wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑛𝑜= {1 if Norway wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑝𝑜 = {1 if Portugal wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝑊𝑖,𝑠𝑝= {1 if Spain wins a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
16
The loss dummies for the specific countries are defined analogously to the win dummies for the specific countries, such that:
𝑋𝐿𝑖,𝑑𝑒= {1 if Denmark loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑒𝑛 = {1 if England loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑓𝑟= {1 if France loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑔𝑒= {1 if Germany loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑖𝑡= {1 if Italy loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑛𝑒= {1 if Netherlands loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑛𝑜= {1 if Norway loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑝𝑜= {1 if Portugal loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑠𝑝= {1 if Spain loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑋𝐿𝑖,𝑠𝑤= {1 if Sweden loses a soccer game for the event that makes 𝑖 the first event after the game 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
As can be seen, the constant term is not included in the regression equation. Due to the fact that all countries are included as dummy variables, including the constant term in the equation would cause a problem of Multicollinearity. This problem is also known as the dummy variable trap. The interpretation of the estimated coefficients on the included country dummy variables will now be that they represent the average value of the dependent variable for each included country. (Brooks, 2008) If we now apply this model to for example to a win by Spain, the equation results in:
𝑅𝑐𝑖 = 𝛽14+ 𝛽1 𝑅𝑐𝑖−1+ 𝛽2 𝑅𝑚𝑖−1+ 𝛽3 𝑅𝑚𝑖+ 𝛽4 𝑅𝑚𝑖+1+ 𝛽5 𝑀𝑖
In this formulation defines 𝛽14 defines the shift of the function for Spain. So due to wins by Spain, the
dependent variable increases on average by 𝛽14. Hypothesis testing can now be used to determine if
there are significant differences between the countries that are included in this regression.
The dummy variables for win and loss games (𝑊𝑐𝑖 and 𝐿𝑐𝑖) are also not included in the regression
17
the changes in the country dummy variables. When the dummy variables for a win and loss are included in the regression, it is not possible to estimate the regression.
5. Results
In this section the results of the ordinary least squares regressions will be presented and discussed. The results for each sub question will be discussed.
The results that are used to give an answer to the first sub question, whether or not the impact of international soccer results on the market returns for Europe as a whole are significant, can be seen in table 5. This is a pooled regression with the 10 countries of the sample. Due to the fact that the residuals of regression 1, 2, and 3 of table 5 are found to be heteroscedastic, these regressions are estimated with heteroscedasticity-robust standard errors.
18
Table 5
The influence of international soccer results on market indices for Europe
This table reports the estimated coefficients and the corresponding p-values of the pooled ordinary least squares regression for the influence of the one day lagged market index return, the one day lagged European index return, the European index return, the European index return plus one day, a dummy variable for the Monday effect, a dummy variable for win games, and a dummy variable for loss games on the market index return of each specific country, for Europe as a whole. Regression 1 is a full regression with all the variables, regression 2 is a regression which accounts for multicollinearity, and regression 3 and 4 are robustness tests.
Dependent Variable Rci
Incl. observations 71712
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob.
β0 0.009% 0.037 0.009% 0.034 0.023% 0.000 0.025% 0.000 Rci−1 -14.636% 0.000 Rmi−1 71.691% 0.000 70.498% 0.000 Rmi 19.444% 0.000 9.146% 0.000 10.686% 0.000 Rmi+1 0.729% 0.108 0.885% 0.059 Mi -0.014% 0.165 -0.020% 0.048 Wci 0.034% 0.499 0.036% 0.496 0.012% 0.863 0.015% 0.838 Lci -0.078% 0.219 -0.063% 0.327 -0.066% 0.453 -0.073% 0.415 R-squared 0.429 0.417 0.009 0.000 Adj. R-squared 0.429 0.417 0.009 0.000 F-statistic 7700.885 8533.800 225.877 0.353 Prob(F-statistic) 0.000 0.000 0.000 0.702
From the matrix of correlations for regression equation 2 (table 6) it can be seen that there is a large correlation (0.639) between the one day lagged index return (𝑅𝑐𝑖−1) and the returns on the European
market index returns (𝑅𝑚𝑖). A solution to multicollinearity is to drop one of the collinear variables. Due
to the fact that 𝑅𝑚𝑖 has a larger impact on 𝑅𝑐𝑖 than 𝑅𝑐𝑖−1, 𝑅𝑐𝑖−1 will be dropped from the regression.
The results of this regression can be seen in regression 2 from table 5. With this regression the dummy variable for the Monday effect (𝑀𝑖) is now significant at a 5% level. Which means that the returns on
Monday are on average 0.02% lower than the other days of the week. The dummy variables for international soccer win and loss games are not significant.
19
Table 6
Matrix of correlations for regression equation 2
This table reports the pair-wise correlations between the explanatory variables of regression equation 2.
Rci Rci−1 Rmi−1 Rmi Rmi+1 Mi Wci Lci
Rci 1 0.023 0.640 0.097 0.016 -0.015 0.001 -0.003 Rci−1 0.023 1 0.088 0.639 -0.028 0.012 0.001 -0.006 Rmi−1 0.640 0.088 1 0.022 0.016 -0.015 -0.002 0.000 Rmi 0.097 0.639 0.022 1 -0.031 0.002 0.002 -0.003 Rmi+1 0.016 -0.028 0.016 -0.031 1 0.008 0.003 0.006 Mi -0.015 0.012 -0.015 0.002 0.008 1 0.047 0.043 Wci 0.001 0.001 -0.002 0.002 0.003 0.047 1 -0.004 Lci -0.003 -0.006 0.000 -0.003 0.006 0.043 -0.004 1
Due to the fact that, for this pooled regression, the p-values of the dummy variables for an international soccer win and loss are insignificant for all regressions, the hypothesis of sub question 1 is false. Therefore, it can be said that there is no significant impact of international soccer results on the market returns for Europe as a whole.
The results for the second sub question, whether or not the market returns on the indices, for each country separate, are significantly affected by international soccer results of that country, can be seen in table 7. Due to the fact that the residuals are found to be heteroscedastic, the regressions are estimated with heteroscedasticity-robust standard errors. Only the coefficient and the significance for the win and loss dummies from the regression which accounted for multicollinearity are shown in table 7.
The regressions and estimates for all the independent variables and the correlation matrix for each country separate can be seen in appendix A.1 until A.20. Due to the fact that for all countries the residuals for regressions 1, 2, and 3, were found to be heteroscedastic, these regressions are estimated with heteroscedasticity-robust standard errors. To summarize the results from the appendix: for the countries Denmark, Germany, and Portugal, the lagged index returns are significant at a 5% level for regression 1. This means that for these countries the returns are influenced by the previous day returns. The variable 𝑅𝑚𝑡, the continuously compounded return on the European market index on day
t, is significant for each country at a 1% level. The returns for the countries of this data sample are thus significantly influenced by European market movements. The dummy variable for the Monday effect is only significant at a 5% level for Italy. This indicates that the returns on Monday in Italy are significantly lower with respect to the other trading days of the week.
From the correlation matrices it can be seen that for all but one country the one day lagged index return (𝑅𝑐𝑡−1) and the one day lagged European market index return (𝑅𝑚𝑡−1) are highly correlated.
20
Table 7
The influence of international soccer results on market indices per country
This table reports the number of included soccer games and ordinary least squares estimates of the coefficients of the win and loss dummy variables with their corresponding p-values for each country that is part of the sample. Each country regression is done with the return of the market index of the specific country as dependent variable, and the following independent variables: one day lagged market index return, the European index return, the European index return plus one day, a dummy variable for the Monday effect, a dummy variable for win games, and a dummy variable for loss games.
Wins (Wct) Losses (Lct)
Country Nr. of games Coefficient Prob. Nr. of games Coefficient Prob.
Denmark 24 0.224% 0.228 19 -0.254% 0.216 England 28 -0.017% 0.860 30 0.001% 0.990 France 32 -0.055% 0.559 18 -0.214% 0.220 Germany 63 -0.046% 0.697 20 -0.141% 0.466 Italy 31 0.073% 0.688 16 -0.077% 0.854 Netherlands 47 0.023% 0.807 33 -0.189% 0.180 Norway 15 -0.280% 0.400 18 0.329% 0.108 Portugal 33 0.034% 0.766 22 -0.201% 0.198 Spain 43 0.014% 0.913 19 -0.314% 0.062 Sweden 22 -0.100% 0.602 25 0.285% 0.161
second regression, because this variable has a smaller impact on the index return 𝑅𝑐𝑡 than the one day
lagged index return (𝑅𝑐𝑡−1). For Spain there was a high correlation between the one day lagged index
return (𝑅𝑐𝑡−1) and the European market index return (𝑅𝑚𝑡). Therefore, the one day lagged index return
(𝑅𝑐𝑡−1) is dropped from the second regression for Spain. The results of these regressions can be seen
in regression 2 from the regression tables of the appendix. This regression which controls for multicollinearity does not provide new insights, other than that the one day lagged index return variable for France is now significant at a 5% level. The dummy variables for an international win or loss game are still insignificant for each country (see table 7).
21
The robustness test which includes only the European market index returns and the international win and loss dummy variables, is as well performed for regression equation 3. The results can be seen in regression 3 and 4 from the regression tables of the appendix. This robustness test does not give new insights. In a regression with only the international win and loss dummy variables as independent variables, there is still no relationship between these dummies and the stock market returns of the countries.
Due to the fact that all, but one, p –values are insignificant from table 7 and that the robustness test does not show new insight, the hypothesis of sub question 2 is false. Therefore it can be said that stock index returns, for each country separate, are not significantly affected by the international soccer results.
22
Table 8
The influence of international soccer results on market indices for southern European Countries This table reports the estimated coefficients and the corresponding p-values of the pooled ordinary least squares regression for the influence of the one day lagged market index return, the one day lagged European index return, the European index return, the European index return plus one day, a dummy variable for the Monday effect, a dummy variable for win games, and a dummy variable for loss games on the market index return of each specific country, for southern European countries. Regression 1 is a full regression with all the variables, regression 2 is a regression which accounts for multicollinearity, and regression 3 and 4 are robustness tests.
Dependent Variable Rci
Incl. observations 18573
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob.
β0 0.004% 0.639 0.006% 0.488 0.009% 0.327 0.016% 0.097 Rci−1 -27.317% 0.000 Rmi−1 54.989% 0.000 46.933% 0.000 Rmi 44.175% 0.000 31.505% 0.000 32.283% 0.000 Rmi+1 1.444% 0.103 2.042% 0.033 Mi -0.028% 0.182 -0.040% 0.068 Wci 0.113% 0.251 0.099% 0.365 0.031% 0.802 0.055% 0.669 Lci -0.138% 0.347 -0.099% 0.525 -0.089% 0.628 -0.139% 0.431 R-squared 0.335 0.281 0.090 0.000 Adjusted R-squared 0.335 0.280 0.090 0.000 F-statistic 1338.936 1207.580 612.613 0.403 Prob(F-statistic) 0.000 0.000 0.000 0.668 Table 9
The influence of international soccer results on market indices for northern European Countries This table reports the estimated coefficients and the corresponding p-values of the pooled ordinary least squares regression for the influence of the one day lagged market index return, the one day lagged European index return, the European index return, the European index return plus one day, a dummy variable for the Monday effect, a dummy variable for win games, and a dummy variable for loss games on the market index return of each specific country, for northern European countries. Regression 1 is a full regression with all the variables, regression 2 is a regression which accounts for multicollinearity, and regression 3 and 4 are robustness tests.
Dependent Variable Rci
Incl. observations 21202
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob.
23
The matrix of correlations for the southern and northern European countries can be seen in table 10 and 11. Both tables reveal high correlations between the one day lagged index return (𝑅𝑐𝑖−1) and the
returns on the European market index returns (𝑅𝑚𝑖). Due to the fact that 𝑅𝑚𝑖 has a larger impact on
𝑅𝑐𝑖 than 𝑅𝑐𝑖−1, 𝑅𝑐𝑖−1 will be dropped from the regression. The results of the regression which accounts
for multicollinearity can be seen in regression 2 from table 8 and 9. With this regression the European index return plus one day (𝑅𝑚𝑖+1) is now significant at a 5% level for the southern countries. This
means that if the European index return plus one day (𝑅𝑚𝑖+1) increases by 1% the market index return
of southern countries will increase by 0.02%. The dummy variables for an international win or loss game are not significant for this regression which accounts for multicollinearity.
Table 10
Matrix of correlations for southern European countries sample
This table reports the pair-wise correlations for southern European countries between the explanatory variables of regression equation 3.
Rci Rci−1 Rmi−1 Rmi Rmi+1 Mi Wci Lci
Rci 1 0.049 0.441 0.300 0.012 -0.014 0.003 -0.006 Rci−1 0.049 1 0.282 0.436 -0.029 0.010 0.003 -0.008 Rmi−1 0.441 0.282 1 0.017 -0.005 -0.002 -0.008 0.002 Rmi 0.300 0.436 0.017 1 -0.016 -0.004 0.005 -0.007 Rmi+1 0.012 -0.029 -0.005 -0.016 1 0.009 0.001 0.009 Mi -0.014 0.010 -0.002 -0.004 0.009 1 0.040 0.048 Wci 0.003 0.003 -0.008 0.005 0.001 0.040 1 -0.004 Lci -0.006 -0.008 0.002 -0.007 0.009 0.048 -0.004 1 Table 11
Matrix of correlations for northern European countries sample
This table reports the pair-wise correlations for northern European countries between the explanatory variables of regression equation 3.
Rci Rci−1 Rmi−1 Rmi Rmi+1 Mi Wci Lci
Rci 1 0.035 0.650 0.051 -0.001 -0.012 0.005 0.007 Rci−1 0.035 1 -0.001 0.650 -0.026 0.018 0.003 -0.007 Rmi−1 0.650 -0.001 1 0.022 0.022 -0.017 0.007 0.004 Rmi 0.051 0.650 0.022 1 -0.038 0.004 -0.003 -0.002 Rmi+1 -0.001 -0.026 0.022 -0.038 1 0.010 0.003 0.015 Mi -0.012 0.018 -0.017 0.004 0.010 1 0.041 0.049 Wci 0.005 0.003 0.007 -0.003 0.003 0.041 1 -0.003 Lci 0.007 -0.007 0.004 -0.002 0.015 0.049 -0.003 1
24
variables, there is still no relationship between these dummies and the stock market returns of southern and northern European countries.
Since the p-values of the dummy variables for an international win and loss for all regressions of both southern and northern countries are not significant, the hypothesis of sub question 3 is false. It can therefore be concluded that the impact of international soccer results is not different for southern and northern European countries.
From the results of the first three sub questions is can be seen that, for this research, there is no evidence that international soccer results of European countries have an impact on the market index returns. The contrast that is found between this research and the findings from the literature review is probably due to differences in the data. This research uses price index returns, because this led to more relevant mood events, but the expectation is that this does not have an impact on the results of this study because possible mood effects can also be seen on a price index, as is explained in the data section (chapter 3). In addition this research only investigates the effect of international soccer results on market indices in European countries and uses a more recent dataset.
The last sub question considers whether there are differences between European countries resulting from the impact of international soccer results on market indices. This sub question can be answered by examining the results from table 12. Due to the fact that the residuals of regression 1, 2, and 3 of table 12 are found to be heteroscedastic, these regressions are estimated with heteroscedasticity-robust standard errors.
25
Table 12
The differences between European countries resulting from the influence of international soccer results on market indices
This table reports the estimated coefficients and the corresponding p-values of the ordinary least squares regression for the influence of the one day lagged market index return, the one day lagged European index return, the European index return, the European index return plus one day, a dummy variable for the Monday effect, dummy variables for win games for the countries, and dummy variables for loss games for the countries on the market index return of each specific country. Regression 1 is a full regression with all the variables, regression 2 is a regression which accounts for multicollinearity, and regression 3 and 4 are robustness tests. From these findings the possible differences in market index sensitivity to win or loss international soccer games between countries can be observed.
Dependent Variable Rci
Incl. observations 71712
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob.
Rci-1 -14.635% 0.000 Rmi-1 71.711% 0.000 70.518% 0.000 Rmi 19.452% 0.000 9.155% 0.000 10.724% 0.000 Rmi+1 0.737% 0.104 0.893% 0.056 Dm -0.005% 0.564 -0.011% 0.229 Xwi,de 0.290% 0.116 0.273% 0.133 0.207% 0.454 0.186% 0.494 Xwi,en -0.022% 0.825 -0.020% 0.835 0.099% 0.606 0.111% 0.658 Xwi,fr -0.046% 0.647 -0.042% 0.674 0.028% 0.867 0.027% 0.909 Xwi,ge -0.041% 0.733 -0.025% 0.838 -0.046% 0.789 -0.042% 0.801 Xwi,it 0.094% 0.633 0.085% 0.669 0.212% 0.513 0.219% 0.359 Xwi,ne -0.007% 0.945 0.000% 0.998 -0.112% 0.465 -0.100% 0.606 Xwi,no -0.229% 0.510 -0.274% 0.416 -0.185% 0.670 -0.163% 0.635 Xwi,po 0.082% 0.534 0.071% 0.568 -0.094% 0.435 -0.078% 0.736 Xwi,sp 0.219% 0.269 0.235% 0.308 0.072% 0.688 0.080% 0.694 Xwi,sw -0.015% 0.933 -0.011% 0.954 0.332% 0.188 0.330% 0.245 Xli,de -0.296% 0.168 -0.268% 0.205 -0.108% 0.650 -0.122% 0.690 Xli,en -0.013% 0.901 -0.003% 0.976 -0.161% 0.395 -0.155% 0.523 Xli,fr -0.287% 0.120 -0.288% 0.121 -0.583% 0.024 -0.595% 0.058 Xli,ge -0.089% 0.639 -0.094% 0.637 0.227% 0.493 0.257% 0.388 Xli,it -0.148% 0.710 -0.136% 0.743 -0.227% 0.614 -0.217% 0.515 Xli,ne -0.190% 0.172 -0.165% 0.241 -0.179% 0.341 -0.200% 0.388 Xli,no 0.267% 0.233 0.313% 0.151 0.530% 0.116 0.562% 0.073 Xli,po -0.243% 0.067 -0.249% 0.067 0.058% 0.832 0.000% 0.999 Xli,sp -0.093% 0.708 -0.050% 0.840 -0.202% 0.393 -0.186% 0.542 Xli,sw 0.308% 0.121 0.312% 0.121 0.234% 0.425 0.218% 0.413 R-squared 0.429 0.417 0.009 0.000 Adjusted R-squared 0.429 0.416 0.009 0.000
26
From the matrix of correlations for regression equation 4 (table 13) it can be seen that there is a large correlation (0.639) between the one day lagged index return (𝑅𝑐𝑖−1) and the returns on the European
market index returns (𝑅𝑚𝑖). Due to the fact that 𝑅𝑚𝑖 has a larger impact on 𝑅𝑐𝑖 than 𝑅𝑐𝑖−1, 𝑅𝑐𝑖−1 will
be dropped from the regression. The results of this regression can be seen in regression 2 from table 12. This regression does not give any new insights. The dummy variables for international soccer win and loss games are not significant.
Table 13
Matrix of correlations for regression equation 4
This table reports the pair-wise correlations between the explanatory variables of regression equation 4. The correlations of the country dummy variables are not included in this table because they are negligible.
Rct Rct-1 Rmt-1 Rmt Rmt+1 Dm Rci 1 0.023 0.640 0.097 0.016 -0.015 Rci-1 0.023 1 0.088 0.639 -0.028 0.012 Rmi-1 0.640 0.088 1 0.022 0.016 -0.015 Rmi 0.097 0.639 0.022 1 -0.031 0.002 Rmi+1 0.016 -0.028 0.016 -0.031 1 0.008 Dm -0.015 0.012 -0.015 0.002 0.008 1
Also a robustness test is performed for regression equation 4, which includes only the European market index returns and the international win and loss country dummy variables. The results can be seen in regression 3 and 4 from the table 12. For a regression with only the European market index returns and the international win and loss country dummy variables (regression 3 from table 12), the loss dummy variable for France is now significant at a 5% level. This can be interpreted as follows: when France losses an international soccer game, the return on the market index following the game will on average be 0.583% lower than returns following no game day. In a regression with only the international win and loss country dummy variables as independent variables, there is no relationship between these dummies and the stock market returns of the countries.
27
6. Conclusion
Behavioural finance uses insights from psychology to understand how human behaviour influences the decisions of individuals and professional investors, markets, and managers. This paper examines the relationship between investor mood and market indices. In particular, it examines the relationship between international soccer results, which is a proxy for investor mood, and stock market returns. Regression analysis, more specific ordinary least squares analysis, is used to investigate whether the international soccer results of 10 European countries have an influence on the market returns of the indices of the corresponding countries.
A dataset containing price index returns and internationals soccer game results, for the 10 European countries of the sample from January 1983 until September 2014 is used to answer the research question:
Are the stock market returns of the leading European indices significantly affected by international soccer game results?
In order to give an answer to this research question, four sub questions had to be answered. For each sub question hypotheses were stated.
There is a body of literature that studies the influence of soccer results on stock returns. Within this field of research there are different types of studies. Some studies examine the influence of the performance of soccer clubs on the stock prices of the specific club, whereas other studies investigate the influence of international soccer results and its effect on local markets. These are the two main types of studies. The main findings from these studies are that national or international win and (especially) loss soccer game results affect the stock prices or the stock market returns.
28
do not show significant influences for both the southern and northern countries. The last sub question, which examines whether the impact of international soccer results on stock market returns is significantly different between European countries is also answered negatively. There are no significant different market index returns between the 10 European countries following a win or loss international soccer game.
It is clear that the results from this study are in contrast with the results found during the literature review. The main reason for this contrast is probably the dataset. This study uses price index returns, European countries, and a recent dataset. The use of price index returns could be a drawback of this research. Price indices are calculated using the price performance of the stock of the index and it completely ignores the dividend pay-outs by companies. The total return index does include these dividend pay-outs while calculating the index performance. As said before, the justification for using price index returns for this research is that for some countries of the sample only a few years of total return index were given. Therefore, using the total return index would result in eliminating a substantial part of the data. However, the expectation is that using the price index returns does not cause a bias for this study because the possible mood effect can also be seen on a price index. The dividends which are reinvested in the total return index are expected to not be affected by international soccer returns, whereas to daily price movements are.
Future research may do this same research for Asia or South-America. Due to the fact that previous studies found that international soccer results do have an impact on market index returns on a global scale, this effect should be recognizable somewhere. According to this research the effect is not recognizable in Europe, so logic would dictate that this effect should be recognizable in other parts (continents) of the world. Another topic for future research might be to investigate the impact of cultural differences on mood, which affects the stock market returns after international soccer games. Even though this study was inconclusive about these cultural differences, a deeper study might reveal that cultural differences do impact mood and therefore stock market returns.
References:
Ackert, Lucy F., and Richard Deaves, 2010, Behavioural Finance. Psychology, Decisions-Making, and
Markets, (South-Western Cencage Learning, Mason).
Ashton, J. K., B. Gerrard, and R. Hudson, 2003, Economic impact of national sporting success: evidence from the London stock exchange, Applied Economics 10, 783-785.
29
Baker, Malcom, and Jeffrey Wurgler, 2007, Investor sentiment in the stock market, Journal of
Economic Perspectives 21, 129-151.
Bell, Adrian R., Chris Brooks, David Matthews, and Charles Sutcliffe, 2012, Over the moon or sick as a parrot? The effects of football results on a club’s share price, Applied Economics 44, 3435-3452.
Benkraiem, Ramzi, Waël Louhichi, and Pierre Marques, 2009, Market reaction to sporting results: The case of European listed football clubs, Management Decision 47, 100-109.
Berthier, Fabrice, and Frédéric Boulay, 2003, Lower myocardial infarction mortality in French men the day France won the 1998 Wold Cup of football, Heart 89, 555-556.
Boido, Claudio, and Antonio Fasano, 2007, Football and mood in Italian stock exchange, ICAFI Journal
of Behavioral Finance 4, 32-50.
Brooks, Chris, 2008, Introductory Econometrics for Finance, (Cambridge University Press, Cambridge). Carroll, Douglas, Shah Ebrahim, Kate Tilling, John Macleod, and George Davey Smith, 2002, Admissions for myocardial infarction and Word Cup football: Database survey, British
Medical Journal 325, 1439-1442.
Demir, Ender, and Hakan Danis, 2011, The effect of performance of soccer clubs on their stock prices: Evidence from Turkey, Emerging Markets Finance & trade 47, 58-70.
Duque, João, and Nuno Abrantes Ferreira, 2005, Explaining share price performance of football clubs listed on the Euronext Lisbon, Working Paper no. 05-01, ISEG-Universidade Tecnica de Lisboa Business Administration.
Edmans, Alex, Diego Garcia, and Oyvind Norli, 2007, Sports sentiment and stock returns, Journal of
Finance 62, 1967-1998.
Engle, Robert F., 1982, Autoregressive conditional heteroskedasticity with estimates of the variance of U.K. inflation, Econometrica 50, 987–1008.
Frieder, Laura and Avanidhar Subrahmanyam, 2004, Nonsecular regularities in returns and volume,
Financial Analyst Journal 60, 29-34.
Hirshleifer, David and Tyler Shumway, 2003, Good day sunshine: Stock return and the weather,
30
Hirt, Edward R., Dolf Zillman, Grant A. Erickson, and Chris Kennedy, 1992, Costs and benefits of allegiance: Changes in fans’ self-ascribed competencies after team victory versus defeat,
Journal of Personality & Social Psychology 63, 724-738.
Hvattum, Lars Magnus, and Halvard Arntzen, 2010, Using ELO ratings for match result prediction in association football, International Journal of Forecasting 26, 460-470.
Kamstra, Mark J., Lisa A. Kramer, and Maurice D. Levi, 2000, Losing sleep at the market: The daylight saving anomaly, American Economic Review 12, 1000-1005.
Kaplanski, Guy, and Haim Levy, 2010, Exploitable predictable irrationality: The FIFA World Cup effect on the U.S. stock market, Journal of financial and quantitative analysis 45, 535-553.
Kolman, Ludek, Niels G. Noorderhaven, Geert Hofstede, Elisabeth Dienes, 2002, Cross-cultural differences in Central Europe, Journal of Managerial Psychology 18, 76-88.
Newbold, Paul, William L. Carlson, and Betty M. Thorne, 2010, Statistics for Business and Economics, (Pearson education, New Jersey).
Palomino, Frederic, Luc Renneboog, and Chendi Zhang, 2009, Information salience, investor sentiment, and stock returns: The case of British soccer betting, Journal of Corporate Finance 15, 368-387.
Qawi, Raluca Bighiu, 2010, Behavioral finance: is investor psyche driving market performance?, Journal
of Behavioral Finance 7, 7-19.
Scholtens, Bert, and Wijtze Peenstra, 2009, Scoring on the stock exchange? The effect of football matches on stock market returns: an event study, Applied Economics 41, 3231-3237.
Stadtmann, Georg, 2003, An Empirical Examination of the news model: The case of Borussia Dortmund GmbH & Co. KGaA, Zeitschrift für Betriebswirtschaft 74, 165-185.
Stracca, Livio, 2004, Behavioral finance and asset prices: Where do we stand? Journal of Economic
Psychology 25, 373-405.
Zuber, Richard A., Patrick Yiu, Reinhold P. Lamb, and John M. Gandar, 2005, Investor-fans? An examination of the performance of publicly traded English premier league teams, Applied
31
Appendix A – OLS regressions for each country separate
These tables will report the estimated coefficients and the p-value of the ordinary least squares regressions and the matrix of correlations for each country. Regression 1 is a full regression with all the variables, regression 2 is a regression which accounts for multicollinearity, and regression 3 and 4 are robustness tests.
Appendix A.1
The influence of international soccer results on market indices for Denmark
Dependent Variable Rct
Incl. observations 6473
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob. β0c 0.018% 0.151 0.018% 0.148 0.018% 0.097 0.030% 0.036 Rct−1 4.092% 0.011 6.383% 0.000 Rmt−1 3.371% 0.023 Rmt 59.948% 0.000 60.037% 0.000 59.949% 0.000 Rmt+1 -1.785% 0.130 -1.852% 0.116 Mt -0.004% 0.882 -0.005% 0.865 Wct 0.230% 0.214 0.224% 0.228 0.221% 0.244 0.156% 0.510 Lct -0.255% 0.215 -0.254% 0.216 -0.278% 0.175 -0.152% 0.567 R-squared 0.413 0.412 0.408 0.000 Adjusted R-squared 0.413 0.412 0.408 0.000 F-statistic 650.225 756.331 1485.208 0.383 Prob(F-statistic) 0.000 0.000 0.000 0.682 Appendix A.2
Matrix of correlations for Denmark
32
Appendix A.3
The influence of international soccer results on market indices for England
Dependent Variable Rct Incl. observations 8279
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob. β0c 0.004% 0.593 0.004% 0.606 0.002% 0.744 0.025% 0.034 Rct−1 -1.304% 0.438 -2.162% 0.062 Rmt−1 -1.010% 0.420 Rmt 73.104% 0.000 73.102% 0.000 73.069% 0.000 Rmt+1 0.852% 0.372 0.867% 0.363 Mt -0.008% 0.654 -0.008% 0.666 Wct -0.017% 0.863 -0.017% 0.860 -0.019% 0.841 0.086% 0.677 Lct 0.002% 0.984 0.001% 0.990 0.000% 0.997 -0.181% 0.364 R-squared 0.621 0.621 0.621 0.000 Adjusted R-squared 0.621 0.621 0.621 0.000 F-statistic 1939.362 2262.430 4515.442 0.501 Prob(F-statistic) 0.000 0.000 0.000 0.606 Appendix A.4
Matrix of correlations for England
33
Appendix A.5
The influence of international soccer results on market indices for France
Dependent Variable Rct
Incl. observations 7100
Independent Regression 1 Regression 2 Regression 3 Regression 4 variables Coef. Prob. Coef. Prob. Coef. Prob. Coef. Prob. β0c 0.007% 0.535 0.007% 0.528 -0.001% 0.894 0.017% 0.311 Rct−1 -3.283% 0.082 -2.436% 0.034 Rmt−1 1.208% 0.525 Rmt 90.801% 0.000 90.806% 0.000 90.781% 0.000 Rmt+1 0.545% 0.617 0.531% 0.625 Mt -0.041% 0.106 -0.041% 0.105 Wct -0.055% 0.560 -0.055% 0.559 -0.067% 0.478 0.010% 0.967 Lct -0.213% 0.223 -0.214% 0.220 -0.222% 0.207 -0.611% 0.061 R-squared 0.638 0.638 0.638 0.000 Adjusted R-squared 0.638 0.638 0.637 0.000 F-statistic 1788.470 2086.474 4160.957 1.761 Prob(F-statistic) 0.000 0.000 0.000 0.172 Appendix A.6
Matrix of correlations for France
