Sports Sentiment and National Stock Markets: Do the win and
loss effect really exist?
12/06/2019
Abstract:
Motivated by previous studies on sports sentiment, this paper investigates whether there exists a relationship between changes in the return on national stock market indices and international football results. Using data from 25 countries at major football tournaments, this study adds value to the ongoing debate on the so-called win and loss effect of football games on stock returns. After controlling for the ability bias between countries, there exists neither a significant relationship between stock returns and wins, nor a relationship between stock returns and losses. Correcting for the time-varying volatility of stock returns using a GARCH(1,1) model does not significantly improve these findings. Robustness checks are performed to check whether the results change when the data is separated based on tournaments, time spans and quality.
JEL Classification: G14, G41, Z23
Keywords: football results, investor sentiment, stock market index, market efficiency
2 1. Introduction
The efficient market hypothesis states that all available information is fully reflected in asset prices (Fama, 1970). Market prices only respond to new information; therefore it is impossible to beat the market on a long-term basis, according to the efficient market hypothesis. Recent literature has studied deviations from the efficient market hypothesis caused by behavioral biases. For example, Hirshleifer and Shumway (2003) found that sunshine is strongly correlated with stock returns. This relationship is caused by upbeat mood. Frieder and Subrahmanyam (2004) confirm that certain religious holidays, like St. Patrick’s Day, lead to abnormal positive returns, while other religious holidays, like Rosh Hashanah, are negatively correlated with stock returns.
This paper will use a different variable for upbeat mood to examine its relationship with a country’s national stock index, namely international football results. A mood variable should satisfy three certain characteristics (Edmans, García and Norli, 2007). Firstly, the variable should drive mood substantially, so the effect is powerful enough to affect the return on assets. Secondly, the variable should impact the mood of a large part of the population to ensure that it covers most investors. Thirdly, the effect should be correlated across most individuals within a certain country. International football results are likely to satisfy all three criteria. Sport results in general have a substantial effect on mood. This is especially the case for football, one of the most popular sports around the globe. Major events like World Cups are likely to take a grip of the majority of a population. According to Wann, McGeorge and Allison (1994), fans react positively to good team performance and negatively to poor team performance. Media coverage shows that football is of ‘national interest’ around the globe. Last year, more than 3.5 billion people watched some official broadcast coverage of the FIFA World Cup 2018 in Russia.1 The World Cup final of 2002 between Germany and Brazil was viewed by more than 1 billion people worldwide. It is hard to think of other events that have an impact on the population’s mood as big as the impact of worldwide football tournaments. While national football results impact the entire country in a similar way, other popular sports, like baseball and American football, are merely played on a club level rather than on country wide level. The characteristics above provide a strong motivation for using football results to examine mood changes of investors and is the key strength of this study.
3 The two main approaches used in previous literature on the topic of interest are the event study approach and the continuous variable approach. There are multiple examples of studies using either an event study approach or a continuous variable approach. Kamstra, Kramer and Levi (2000) used an event study approach to examine the effect of changes in daylight saving on the disruption of sleeping patterns. Three years later they analyzed daylight with a continuous variable approach (Kamstra, Kramer and Levi, 2003). The main advantage of the event study approach, rather than the continuous variable approach, is that the former gives a high signal-to-noise ratio in returns, because sudden changes in investor’s mood can be identified clearly. The main disadvantage of the event study approach is that the number of observed signals often appears to be low. In this study, a continuous variable approach is used. Some papers that used the same approach are Edmans et al. (2007) and Fan and Wang (2018). The reason that the continuous variable approach is preferred over an event study approach is that football wins and losses occur often, and not on a single event basis. When an event study approach is used, there needs to be a discrete event. In the case of sports sentiment, this could be the winning of the final at a major football tournament. In this study, not only these finals are taken into account, but all other matches during major football tournaments as well. This leads to a total of 1361 games. Kamstra et al. (2000) preferred an event study approach, because changes in daylight saving only occurs twice a year, once close to the start of spring and once in autumn. Since hundreds of football games are played each year by national teams, it is hard to distinguish single events and the event study approach is not very applicable.
4 a game against Croatia and a game against Iceland. Both losses are indicated as unexpected, because England is higher on the FIFA ranking than both Croatia and Iceland. However, the surprise after the loss against Iceland should be more pronounced, because the difference in FIFA ranking between England and Iceland is larger than the difference in FIFA ranking between England and Croatia. Later in this paper will be explained how the different levels of surprise are measured. Lastly, several robustness checks are added to check whether the results are sensitive to changes in tournaments, time spans and quality.
Although most previous studies find a significant relationship between stock returns and football results, there are also studies that contradict these findings. The literature review outlines these different findings of previous papers. The fact that most of the previous studies use data before 2006 emphasizes the importance of this new study, using newly available data from football results until 2018. A follow-up study of Ashton, Gerrard and Hudson (2011) extended its previously used dataset and found that over the last period the effect of football results on stock results decreased. It will be interesting to examine whether this trend has continued during the last ten years.
The research question in this paper reads: ‘‘How do national stock indices respond to changes in investor’s mood caused by international football results?’’ The hypotheses are that there exists a positive relationship between returns on stock markets and international football wins, and a negative relationship between returns on stock markets and international football loses. The rest of the paper is organized as follows. Section 2 discusses previous studies on the relationship between stock returns and sports sentiment. Section 3 shows from where the data is gathered, and which research method is used in this study. Section 4 gives the econometric results and section 5 concludes the paper.
2. Literature review
5 market hypothesis, these deviations from the efficient market hypothesis will be discussed in the following literature review.
I. The loss effect
A diverse range of studies on football results and stock returns have been published prior to this paper. Most of the researchers found a positive relationship between stock returns and wins, and a negative relationship between stock returns and losses. Edmans et al. (2007) find a significant loss effect, indicating the negative relationship between stock returns and losses. However, the effect of wins on the return on the national stock market is insignificant. Apparently, the effect of sports results on investor’s upbeat mood is smaller than the effect on investor’s negative mood. Psychological literature shows significant differences in behaviour by a country’s population after wins and after losses. According to Kloner, McDonald, Leeka and Poole (2009), losses lead to an increase in crimes, suicides and heart attacks, while there is no significant mood improvement after a positive sport result.
This is in line with the prospect theory of Kahneman and Tversky (1979). This theory tells that the main driver of utility is carried by changes in utility (gains and losses) measured from a certain reference point, rather than wealth levels. According to Markman and Hirt (2002), fans are subject to an ‘allegiance bias’, that tells that individuals who are psychologically invested in a desired outcome have biased predictions. If fans predict a victory, losses will have a larger effect on mood than wins. Additionally, in knockout tournaments, a loss immediately leads to abrupt elimination from the tournament, while a win only ensures a spot in the next round. All this explains the existing loss effect and the insignificant effect for wins.
6 hypothesis, the economic impact of the expected game result should be rationally included in the price prior to the game. The loss effect should therefore be larger for losses that are unexpected. The finding of Fan and Wang (2018) is that there is no significant impact of unexpected outcomes on local stock returns.
In other sports, like cricket, basketball and rugby, there also appears to be significant evidence for a loss effect. However, the magnitude of the relationship for these sports is smaller than for football, explained by the phenomenon that football is the most important sport in most countries. By performing robustness checks, Edmans et al. (2007) found that the magnitude of the effect increases for smaller stocks and more important games, like World Cup finals. The larger effect for more important games is in line with Fan and Wang (2018). They found an increasing effect on stock markets during rivalry games, which can be labelled as more important games. Sports rivalry lead to intense competition between teams and especially supporters. In the context of football games of national teams, the Netherlands against Germany or Brazil against Argentina are examples of rivalry games. These nations are close to each other geographically, and have a long history. Media attention can mitigate the effect of rivalry games. Fan and Wang (2018) give two possible explanations for this. More media discussions will increase the number of scenarios fans will have in their mind and no result is therefore unexpected. Furthermore, media attention diminishes the winning effect, since the winning side might feel relieved after the game due to all the tension built up prior to the game. Fan and Wang (2018) contradict Edmans et al. (2007) on the win effect. While Edmans et al. (2007) did not find a significant win effect, Fan and Wang (2018) find the opposite result that winning rivalry games significantly boosts investor’s confidence. For some teams, winning a rivalry game can completely offset a poor season. Even though this is especially true for club teams, this can be the case for national teams as well.
II. Different sentiments and critique on the loss effect
7 foreign investors, because the financial sector is one of the most liquid sectors. According to Kaplanski and Levy (2010), 33% of all transactions on the U.S. stock market in 2006 were conducted by foreign investors. Foreign investors are expected to be more prone to football sentiment, because football appears not to be a very popular sport among local U.S. investors (Curatola et al., 2016). The findings of Curatola et al. (2016) are consistent with the observation of Dahlquist and Robbertson (2001) that foreign investors prefer more liquid stocks. Money can be earned by short selling during the World Cup and buying back stocks afterwards. Curatola et al. (2016) claim to know the behavioural argument behind the negative relationship between stock returns and losses. Investors will enter a state of bad mood when the nation they are supporting loses and will decide to sell its stocks. Consequently, the return on the stock will fall.
Ashton et al. (2003) introduce a feelgood factor to relate the performance of the English national team at World Cups to changes in the FTSE 100. A win increases the confidence for the future (feelgood factor), and there are potential economic benefits from getting far through the tournament. This explains the findings from most studies that the relationship between stock returns and results is smaller for friendly games.
8 German loss would lead to a result of 10 times the stake. Surprisingly, Germany lost, got eliminated from the European Championship, and experienced severe media responses. Rather than using betting odds, Edmans et al. (2007) used Elo Ratings in the same model, giving national football teams ratings based on success in previous games. However, previous results are not necessarily reflected in expectations on current games, hence Klein et al. (2009) prefer betting odds. Both econometric frameworks that are used do not show a significant relationship between positive (negative) stock returns and wins (losses). In both models, the national stock index of Denmark does even fall when the Danish national team wins, a counterintuitive finding. Even in the model that controls for surprising results, Klein et al. (2009) do not find a significant relationship and question the findings of Ashton et al. (2003). The few significant relationships between stock indices and football results that are found by Klein et al. (2009) also differ with respect to the employed event study framework and are therefore likely to be model-driven or a coincidence. Klein et al. (2009) are absolutely sure that they could have found significant results if they would have enlarged the datasets and included many variables. They emphasize that the significant relationship found by Ashton et al. (2003) is probably a ‘sheer coincidence’ and include the term ‘publication bias’, further examined by Geyer-Klingeberg, Hang, Walter and Rathgeber (2018).
9 point is that his/her team/nation will win, leading to a stronger reaction after a loss than after a win.
As a response on the critique of Klein et al. (2009), Ashton et al. (2011) address the four pitfalls by adjusting the model and extend the dataset to 2009 instead of 2002. They still find a significant relationship between stock returns and football results. However, the effect on stock returns did decrease over the last period, from 2002 to 2009. Ashton et al. (2011) state that the interpretation of results by Klein et al. (2009) are strongly biased towards wishing to support the efficient market hypothesis, by arguing that Klein et al. (2009) downplay the importance of its significant findings.
III. Publicly traded clubs, fanaticism, risk tolerance and inattention
Instead of using results of national football teams, Floros (2014) examines the effect of draws (ties) of four European football clubs (Ajax, Benfica, Juventus and Porto) in the Champions League and UEFA Cup on the club’s own stock price. For Benfica and Ajax, draws have a positive effect on its stock price. Draws of Juventus have a negative effect on the stock price and no effect is found for Porto. Juventus is the only club for which losses have a negative effect on the stock return. For the other football clubs, investors take a ‘hold’ position after losses until the next match. The different findings between the clubs are likely to be explained by the size of the clubs and the expectations of winning. Apparently, draws are bad news for Juventus, good news for Ajax and Benfica, while investors in Porto take a ‘neutral’ position. One pitfall is that Floros (2014) does not find a significant relationship between stock returns and wins. Floros (2014) also examine a stylized fact of financial volatility, that bad news (falling market) has a larger effect on volatility than good news (rising market). In this specific setting, bad football results are expected to drive down stock returns, causing an increase in the leverage of the stock and leading to a more volatile stock.
10 Berument, Ceylan and Ogut-Eker (2009) include fanaticism to find the effect of international wins of Turkey’s largest football clubs (Besiktas, Galatasaray and Fenerbahçe) on the Istanbul Stock Exchange. The level of fanatism is based on the intensity of cheering, policy reports on crime rates prior to the game and questionnaires filled in by fans. International games are used, rather than domestic games, because domestic results can be positive for one team and negative for the other. This can possibly lead to the effect to cancel out. Additionally, Turkish wins in international games can create positive spillover effects for domestic non-fans of the team. In line with most previous literature, Berument et al. (2009) find a negative relationship between stock returns and football results after losses, and a positive relationship after wins. The positive relationship between the return on the stock index and wins is strongest for Besiktas, the club with the highest level of fanaticism.
Berument and Ceylan (2013) used the same dataset to relate football results to risk tolerance. In this study, the conditional variance of the stock market return is used as a risk measure. After a football team’s win, agents discount future events more favourably and become more tolerant towards risk. Similarly, after a loss, the agents are expected to be less tolerant towards risk. However, results from the study on the three Turkish teams show that the relationship is only significant for wins and not for losses. This is partly in line with findings of their study a year before. Specifically, Berument and Ceylan (2012) study international cup results from clubs from four countries where football is the nation’s most important sport and influences the daily life of the country’s population. Berument and Ceylan (2012) distinguish football giants from Spain and Britain from relative less successful football clubs from Chili and Turkey. The main finding is that losses of football giants lead significantly to lower risk tolerance and wins of the less successful nations lead to significant higher risk tolerance. The explanation for this is that if a certain result is unexpected, the impact of it will be bigger. This explanation is in line with Edmans et al. (2007). Losses by Spanish and British teams are unexpected and significantly affect risk tolerance, as well as wins of Chilean and Turkish teams.
11 price formation. The MSCI World Index can be reduced by up to 21% during football matches of national teams. Both the trade volume and the stock return convert back to its ‘normal’ value after around 30 to 60 minutes after the match. It has to be noted that the exact change in stock returns is also influenced by the score on the pitch.
A few years after their paper on inattention, Ehrmann and Jansen (2016) found that a country’s stock index is underpriced by up to 7 basis points when they are behind in the game. When elimination becomes more likely, the effect of underpricing increases. By investigating live stock returns, Ehrmann and Jansen (2016) shed a new light on previous literature, that often only focusses on stock returns the day after the game. Ehrmann and Jansen (2012) conclude that stock markets follow developments on the pitch rather than developments in the trading world. Investors should be aware that market liquidity strongly fluctuates when the national team is playing. This supports the view in previous studies on behavioural finance that investors can act irrational.
IV. Implications literature review
The literature review above clearly shows the disagreement between previous studies on the relationship of stock returns and football results. Some studies found both a win and loss effect, some studies found either a win or a loss effect, and some studies found no effect at all. By analysing the literature review, this study can build on the work that is done in previous studies. It is useful to see the differences in research methods in different studies. Some studies included risk tolerance, while a different study included fanaticism to find a relationship between stock returns and football results. Combining new data with different variables (surprise effect), this paper can shed a new light on possible violations of the efficient market hypothesis.
3. Data and methodology
I. Data sources and restrictions
12 are correct, 100 results from different years are randomly checked using websites of the organization of the tournaments.2 It appears that the games in the dataset are correct. The data is cleaned by setting several restrictions. Only the World Cup, the five continental cups (European Championship, CONCACAF Gold Cup, AFC Asian Cup, Copa América and Africa Cup of Nations) and the Confederations Cup, a tournament a year prior to the World Cup are included in the dataset. Only these tournaments are included, because these tournaments have the largest media attention and are most likely to affect the mood of individuals. Matches before 1999 are excluded, below will be explained why this is the case. This leads to a time span of 20 years, from 1999 to 2018. Only the 25 countries that played the most matches are included. Mexico played the most matches (117 matches) and the countries included with the fewest matches played are Ecuador and Saudi Arabia (both 34 matches). To get a reliable and consistent result, it is chosen that countries that did play too few games are excluded. Also, the games that are included in this model still cover a major part of the total matches played during the main football tournaments. The selection effect might be a possible concern if only countries with the most matches played are included in the dataset. A critique can be that countries that played more games are better football nations and are more likely to win. However, this concern is not an issue since table 1 shows that 672 out of the 1361 total games are wins. Although there are more wins than losses, the difference between the number of wins and losses is not large. Besides, the 25 selected countries also consist of countries that are far from football giants. Countries like Australia, Saudi Arabia, Peru and Tunisia succeed to qualify for many tournaments, but are not expected to be strong contenders during the tournament. According to the FIFA World Rankings (table A3 in the appendix), these three countries are ranked on average 48th, 54th and 55th, respectively. Later in this paper, more attention will be paid to the differences in strength between football nations. Firstly, differences in ability will be included in the model using a control variable that takes different values for different level of surprises. Besides, the countries will be split up as ‘‘football giants’’ and ‘‘football dwarfs’’ as a robustness check.
Some countries, like Cameroon, Senegal and Uruguay, do not have a national stock market index and could not be used. In all countries that were left, football is one of the most popular sports, which is a necessary condition to have an effect on investor’s mood.3 After the data
cleaning, the dataset contains 25 countries with a total of 1361 observations of wins and losses.
13 Data on the value of the stock indices is gathered from Thomson Reuters Eikon. Data on the stock indices from this source is available from February 1999 onwards, that is why football games before this year are excluded. Returns can be found by comparing the stock market rate the day after the game to its value the (trading) day before. The main stock index in Colombia, the IGBC index, was introduced after 1999. Therefore, the games played by Colombia before the introduction of its index in 2001 had to be deleted. This is not an issue, since there are still 51 games of the Colombian national team left in the dataset. Table A1 in the Appendix states the stock market indices that are used per country. The most important stock market index in each country is chosen as the relevant index.
Table 1: Summary statistics of game day return per country
Note: the mean daily return is measured as the mean daily return after matchdays, not the mean of all days throughout the year.
Table 1 above gives the summary statistics of the stock market return the day after the game for each country and all countries combined. As expected, the mean post-game daily return of Country Wins Draws Losses
14 all countries combined after losses is lower than the mean return after wins. However, the difference between the mean returns is very small. Surprisingly, not only the mean return for losing is negative (-0.12%), but also the mean return after a win is negative (-0.10%). The mean post-game daily return after draws is neither negative nor positive (0.00%). The mean return after draws for top football nations is in general lower than the mean return for smaller football nations. For example, the mean returns of Argentina and Italy are -0.84% and -0.68%, respectively. This is in line with expectations, because a draw for football giants can feel like a loss. However, one needs to be careful when interpreting these mean returns, since there are not that many draws observed as wins and losses.
II. Benchmark model
To find a relationship between the return on stock market indices and football results, two models are estimated. The residuals from the first model are saved to find the correct coefficients of the second model.
The first model that is estimated is as follows:
𝑅𝑖𝑡 = 𝛽0𝑖+ 𝛽1𝑖𝑅𝑖𝑡−1+ 𝛽2𝑖𝑅𝑚𝑡−1+ 𝛽3𝑖𝑅𝑚𝑡+ 𝛽4𝑖𝑅𝑚𝑡+1+ 𝛽5𝑖𝑀𝑡+ 𝛽6𝑖𝐻𝑡+ 𝛽7𝑖𝑆𝑖𝑡+ 𝜀𝑖𝑡 (1)
𝑅𝑖𝑡 is the daily return on the national stock index of a specific country i on day t, the day after the country played a game. The lagged return of the national stock index 𝑅𝑖𝑡−1 is included as well to account for first-order autocorrelation. The daily U.S. dollar return on the world market index (MSCI All Country World Index) 𝑅𝑚𝑡 is included, to account for integration of international stock markets, which causes correlation of national indices across countries. Some national stock indices are leading the world market and some indices are lagging the world market, therefore 𝑅𝑚𝑡−1 and 𝑅𝑚𝑡+1 are included. The view of leading and lagging markets is supported by a lot of recent literature. For example, Copeland and Copeland (1998) find that the United States has statistically significant one-day leads over world markets, while the Canadian TSX Composite Index is lagging the MSCI World Index according to charts of Dynamic Tree Asset Management (2017).
𝑀𝑡 = {𝑚1𝑡, 𝑚2𝑡, 𝑚3𝑡, 𝑚4𝑡} includes four dummy variables for Monday through Thursday to
15 literature. The individual return on Monday is often higher than the return on other trading days. If the Monday effect is not controlled for, the relationship between stock returns and sports sentiment is likely to be overestimated. Table 2 shows on what days of the week the matches in the dataset are played most. Although the number of games per day is reasonably divided, still most games are played during the weekend. The daily return of 675 games, almost half of the total games, are measured on the first trading day after the weekend (Monday). This emphasizes the importance of controlling for the Monday effect. If a game is played on Friday, Saturday or Sunday, 𝑚1𝑡 takes the value 1. If a game is played on Monday, 𝑚2𝑡 takes the value
one. Stock returns after games on Tuesday are measured on Wednesday, for these games 𝑚3𝑡 takes the value one. 𝑚4𝑡 takes the value one for games on Wednesday and all four dummy variables are zero for games played on Thursday.
Table 2: Number of games played per day of the week
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
145 168 203 170 146 255 274
Total games played: 1361
𝐻𝑡 = {ℎ1𝑡, ℎ2𝑡, ℎ3𝑡, ℎ4𝑡, ℎ5𝑡} are dummy variables for which the previous 1 through 5 days are non-weekend (public) holidays. According to Frieder and Subrahmanyam (2004), (public) holidays are expected to have an effect on stock returns. They show that stock returns are significantly higher on days preceding non-weekend holidays. Possible reasons are that investors want to cover a short position before closed-market days and the existing positive sentiments prior to festive days. The impact of sport results on stock returns is likely to be overestimated if the ‘holiday effect’ is not taken into account. If the day before t is a non-weekend holiday, ℎ1𝑡 takes the value one, otherwise it is zero. If the two day before t is a non-weekend holiday, ℎ2𝑡 takes the value one, otherwise it is zero. The other three dummy variables
can be interpreted in the same way.4
The final control variable that is included in the first model is a variable that controls for ability (𝑆𝑖𝑡). Not many previous studies paid attention to the differences in ability between countries.
A World Cup loss for a relatively weak football nation like Nigeria is less harsh than a World Cup loss for a football giant like Spain. A win of a weak football nation and a loss of a strong
16 football nation are expected to have a stronger effect on the investor’s sentiment. The win probability of weak football nations is lower than the win probability of strong football nations. Therefore, the model needs to control for this ability bias. In 1992 the FIFA introduced the FIFA World Rankings; a ranking system to measure the ability of football nations in absolute values. Using the FIFA World Rankings, differences in ability of football nations can be easily examined. By basing the value of the ranking on present and past performances, the FIFA World Rankings give a fair and clear view on the strength of a football nation. The ranking is updated every month, so changes in ability are quickly adjusted to current results. When studying national football teams, it is important that changes in ability are quickly adjusted, because the strength of national football teams is likely to vary over time (Koning and McHale, 2012). This is caused by two factors. Firstly, a national team can only select players from that specific nation. It cannot use money to hire high-quality players from abroad. A national team runs the risk that it has to use a group of poor-quality players for multiple years. In contrast, top clubs like Real Madrid and Manchester City can use their immense budgets to ensure the strength of its squad to be constant. Secondly, national teams play fewer games than football clubs. This second factor strengthens the effect of the first factor. For the reasons above, in this model the FIFA World Rankings are used to account for the ability bias. Like the data on the football results, data of the FIFA World Rankings is collected from Kaggle.com as well. For each game in this study, the FIFA rankings at the moment the game was played are compared between the two opponents. If the team with the lower (higher) FIFA ranking wins (loses), this can be seen as a surprise. However, when the difference in FIFA rankings of the two opponents is small, a draw could be more expected than a win or a loss. Besides, the surprise is expected to be larger if a strong football nation loses against a very poor nation than after a loss against a team that is neither strong nor poor. Therefore, the relative FIFA ranking (𝐹𝑅𝑖
𝐹𝑅𝑗) is
17 effect (𝑆𝑖𝑡) for each game can be measured by subtracting the actual points gathered (𝑃𝑖𝑡) by the expected number of points (𝐸[𝑃𝑖𝑡]). This is indicated by equations (2) and (3) below. 𝐸[𝑃𝑖𝑡] = 3 ∙ 𝑝𝑤 + 1 ∙ 𝑝𝑑 + 0 ∙ 𝑝𝑙 (2)
𝑆𝑖𝑡 = 𝑃𝑖𝑡− 𝐸[𝑃𝑖𝑡] (3)
where 𝑝𝑤 is the proportion of wins per subgroup, 𝑝𝑑 the proportion of draws per subgroup and 𝑝𝑙 the proportion of losses per subgroup.
Table 3: Results and expected points per subgroup, sorted on relative FIFA World Rankings
When analysing the expected result of a game, it is important to take into account the home advantage. There is extensive literature on the home advantage in football that provide several explanations for this phenomenon. Some of the reasons given by Pollard (2008) are the crowd effects, travel effects, referee bias and territoriality. With no home advantage, the number of points gained by home teams as a percentage of total points earned in a league would be 50%. The higher the percentage above 50%, the higher the influence of the home advantage.
Pollard and Gómez (2014) examined this ratio for 157 football nations and found an average ratio of 57.3%. Therefore, the average home advantage factor is 1.146. The expected number of points (𝐸[𝑃𝑖𝑡]) in equation (2) should be adjusted when country i hosts a tournament and experiences home advantage (equation (4)). Besides, when country i plays against a country that hosts a tournament, the expected number of points should be reconsidered. Equation (5) states the expected number of points of country i when its opponent (country j ) hosts one of the football tournaments.
Subgroup (𝐹𝑅𝑖
𝐹𝑅𝑗) Wins Draws Losses Expected points
18 𝐸[𝑃𝑖𝑡] = 1.146 (3 ∙ 𝑝𝑤+ 1 ∙ 𝑝𝑑 + 0 ∙ 𝑝𝑙), 𝑤ℎ𝑒𝑛 𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑖 𝑖𝑠 ℎ𝑜𝑠𝑡 (4) 𝐸[𝑃𝑖𝑡] =3∙𝑝𝑤+1∙𝑝𝑑+0∙𝑝𝑙
1.146 , 𝑤ℎ𝑒𝑛 𝑐𝑜𝑢𝑛𝑡𝑟𝑦 𝑗 𝑖𝑠 ℎ𝑜𝑠𝑡 (5)
As an example, take the FIFA World Cup 2010 in South Africa. On the 22nd of June 2010,
South Africa surprised the whole world with a 2-1 win against France. At that moment, the ratio of FIFA World Rankings points of South Africa to France was 0.35. Without home advantage, the expected number of points for South Africa would be 0.9 and the surprise effect would be equal to 2.1. However, South Africa’s home advantage led to an expected number of points of 1.03 (0.9·1.146) and a surprise effect of 1.97. In this case, the home advantage lessens the surprise effect.
In the second model (equation 6), the possible win and loss effects are estimated. The residuals from regression (1) are used as abnormal return 𝜀̂𝑖𝑡 to find the relationship between stock returns and match outcomes as follows:
𝜀̂𝑖𝑡 = 𝛼𝑤𝑊𝑖𝑡 + 𝛼𝑙𝐿𝑖𝑡+ 𝑢𝑖𝑡 (6) 𝑊𝑖𝑡 is a dummy variable that takes the value one if country i has won a specific match and takes the value zero if the country has lost. Vice versa, 𝐿𝑖𝑡 is a dummy variable that takes the
value zero if the country has won and takes the value one if country i has lost a game. When the game has ended in a draw, both the win and loss variables are equal to zero. Since the win and loss dummy variables are multicollinear, one of the two variables will be omitted if the model includes a constant. To prevent this, the constant is removed.
The null hypothesis (equation 7) is that football results do not affect stock markets and hence, that the efficient market hypothesis holds. In other words, this states that the win and loss variable of the second model (equation 6) are not significantly different from zero.
𝐻0: 𝛼𝑤 = 0 and 𝛼𝑙= 0 (7)
The alternative hypothesis (equation 8) is that there exists a positive relationship between stock market returns and wins, and that losses lead to a negative reaction in the national stock market. The alternative hypothesis is supported by most previous literature.
19
III. Time-varying volatility
The benchmark model above is based on the assumption of constant volatility. However, it is not uncommon that the volatility of stock returns varies over time. The standard errors can be biased if games during high volatility periods are compared to games played in periods with low volatility. Figure 1 plots the daily return on the MSCI All Country World Index (the World Market Index) over the period that is examined (1999-2018). It clearly shows that the volatility of the stock returns is not constant over time. This indicates the importance of controlling for the time-varying volatility. The same reasoning applies for the stock market index of all 25 countries in this study. The figures with the daily returns of these countries can be found in the appendix (Figure A1-A25).
Figure 1: Daily return MSCI All Country World Index (1999-2018)
To take into account the time-varying volatility, a GARCH(1,1) model is used. This model was originally proposed by Bollerslev (1986). This model was included in the paper of Edmans et al. (2007) as well. The GARCH(1,1) model is preferred over an ARCH model, because it is more parsimonious and it requires less parameters. Higher orders are not necessary, because in almost all cases a GARCH(1,1) model is sufficient to capture all the volatility clustering. After running the regression in equation (3), the volatility of the error term is modelled as follows: 𝜎𝑖𝑡2 = 𝛾0𝑖+ 𝛾1𝑖𝜀𝑖𝑡−12 + 𝛾2𝑖𝜎𝑖𝑡−12 (9)
In this GARCH(1,1) model, the conditional variance at time t does not only depend on the lagged squared error term at time t-1, but also on the lagged variance at time t-1. 𝜎𝑖𝑡2 is the volatility of 𝑅𝑖𝑡, the daily return on the national stock index.
20 Next, the stock returns 𝑅𝑖𝑡 are normalized into 𝑅𝑖𝑡∗ by using 𝜎̂𝑖𝑡 as follows:
𝑅𝑖𝑡∗ = 𝑎𝑖 + 𝑏𝑖 (1
𝜎̂𝑖𝑡) 𝑅𝑖𝑡 (10)
𝑎𝑖 and 𝑏𝑖 are chosen such that the mean of 𝑅𝑖𝑡∗ is zero and its variance is one. By normalizing the returns on the national stock market index of all 25 countries, the heterogeneity in volatility across countries is eliminated. The model in equation (1) is modified using normalized returns 𝑅𝑖𝑡∗ rather than 𝑅𝑖𝑡. In the same way, the lagged return (𝑅𝑖𝑡−1), the return on the world market
index (𝑅𝑚𝑡) and both its lagged and leading variables (𝑅𝑚𝑡−1 and 𝑅𝑚𝑡+1) are normalized. The
normalized residuals 𝜀̃𝑖𝑡 that are obtained from the new model are denoted as abnormal normalized returns and modify equation (6). The equations will be modified as follows: 𝑅𝑖𝑡∗ = 𝛽0𝑖+ 𝛽1𝑖𝑅𝑖𝑡−1∗ + 𝛽2𝑖𝑅𝑚𝑡−1∗ + 𝛽3𝑖𝑅𝑚𝑡∗ + 𝛽4𝑖𝑅𝑚𝑡+1∗ + 𝛽5𝑖𝑀𝑡+ 𝛽6𝑖𝐻𝑡+ 𝛽7𝑖𝐴𝑡+ 𝜀𝑖𝑡 (11) 𝜀̃𝑖𝑡 = 𝛼𝑤𝑊𝑖𝑡+ 𝛼𝑙𝐿𝑖𝑡 + 𝑢𝑖𝑡 (12)
4. Results
I. Win and loss effect per period
21 increase of 1.37 percentage point on the daily stock market return during the period 1999-2018. However, these coefficients are also not significant and therefore one needs to be careful when interpreting the coefficient.
Table 3: Win and loss effect per period
Significance: * denotes p < 10%, ** denotes p < 5% and *** denotes p < 1%
II. Win and loss effect per tournament
As a robustness check, the data is divided per tournament. There are differences in importance present between tournaments. For example, during the FIFA World Cup there is a lot more at stake than during the Confederations Cup, that simply serves as a preparation tournament a year prior to the FIFA World Cup. There are also differences expected between continental cups. For example, the Copa América is played in South America where people react in general more emotional than people elsewhere in the world. The probability that a win or loss effect is present in countries with more fanaticism is expected to be higher than in countries with less fanaticism. Berument et al. (2009) found the strongest relationship between stock returns and wins for Besiktas, the club with the highest level of fanaticism in Turkey.
Table 4 shows that none of the seven tournaments have a win or loss effect that is significantly different from zero. The loss variable of the European Championship (EURO) from the GARCH model is the closest one to being statistically significant with a t-value of -0.58 and a corresponding p-value of 0.561. This value is still far away from being significant at the 10% significance level. For some tournaments it even appears that the coefficient for wins is negative
Wins Losses Number of wins 𝛼𝑤 t-value Number of losses 𝛼𝑙 t-value I. Abnormal returns All games 672 0.0002976 0.69 415 -0.0005621 -1.03 1999 – 2005 247 0.0001222 0.16 165 -0.0006559 -0.70 2006 – 2018 425 0.000517 1.03 250 -0.0005699 -0.87
II. Abnormal normalized returns
All games 672 0.0137305 0.44 415 -0.0281949 -0.72
22 and the coefficient for losses is positive, an unexpected outcome. The t-values corresponding to these coefficients are nonetheless relatively close to zero as well.
Table 4: Win and loss effect per tournament
Significance: * denotes p < 10%, ** denotes p < 5% and *** denotes p < 1%
III. Win and loss effect for football giants and football dwarfs
As a final robustness check, the data is divided based on quality of the football nations. Based on the average FIFA ranking over the period 1999-2018, the football nations can be compared in terms of quality. The best thirteen countries are indicated as ‘football giants’ and the worst twelve countries are called ‘football dwarfs.’ Table A3 in the appendix states which countries belong to which subgroup and what the average FIFA ranking of the corresponding country is. In an earlier study, Berument and Ceylan (2012) distinguished football giants from football dwarfs to find a relationship between risk tolerance and football results. According to that study, unexpected results lead to a significant change in risk tolerance. Losses are unexpected results for football giants, while wins are unexpected results for football dwarfs. Berument and Ceylan (2012) did not study the relationship between stock returns and football results. This makes it
Wins Losses Number of wins 𝛼𝑤 t-value Number of losses 𝛼𝑙 t-value I. Abnormal returns World Cup 169 0.0002932 0.31 130 0.0000336 0.03 Africa Cup 69 -0.0005079 -0.37 32 0.000055 0.03 Asia Cup 51 0.0001185 0.07 19 -0.0014773 -0.53 Copa América 108 0.0001796 0.19 81 0.0001622 0.15 EURO 78 -0.0001832 -0.21 39 -0.0004926 -0.40 Gold Cup 114 -0.0000176 -0.03 51 0.0001981 0.23 Confed. Cup 83 0.0002123 0.22 63 -0.0002797 -0.26
II. Abnormal normalized returns
23 interesting to see whether the robustness check in this study gives new insights. The results from the robustness check are stated in table 5. For both football giants and football dwarfs the win coefficients are positive, and the loss coefficients are negative. Although none of the coefficients are significant again, the coefficients for the win and loss effect of football dwarfs are relatively close to being significant with t-values of 0.99 and -0.97, respectively. Additionally, the magnitude of the win and loss coefficient are stronger for football dwarfs than for football giants. An explanation for this can be the difference in financial literacy between the countries that are classified as football giants and the countries that are classified as football dwarfs. According to a report of Klapper, Lusardi and Van Oudheusden (2015), the average financial literacy of the football dwarfs is smaller than the average financial literacy of the football giants. Investors from weaker football nations might therefore be more prone to investor sentiment than investors from stronger football nations. However, the insignificant coefficients cause these interpretations to be risky.
Table 5: Win and loss effect for football giants and football dwarfs
Note: Table A3 in the appendix states which countries are classified as ‘football giants’ and which countries are classified as ‘football dwarfs.’ Significance: * denotes p < 10%, ** denotes p < 5% and *** denotes p < 1%
It appears that for all set of results, the coefficients are insignificant. After controlling for different periods, different tournaments and different quality of football nations, the win and loss effect are still not present. Apparently, the findings of Klein et al. (2009) are well-grounded. As is discussed in section 2, the win and loss coefficients for almost all countries in the paper of Klein et al. (2009) were insignificant. The very few coefficients that appear to be significant are counterintuitive: losses are followed by positive returns and wins by negative returns. This
Wins Losses Number of wins 𝛼𝑤 t-value Number of losses 𝛼𝑙 t-value I. Abnormal returns Giants 456 0.000246 0.50 204 -0.000485 -0.66 Dwarfs 216 0.0007436 0.99 211 -0.0007382 -0.97
II. Abnormal normalized returns
Giants 456 0.0166378 0.48 204 -0.0277558 -0.53
24 study supports the critique of Klein et al. (2009) on the paper of Ashton et al. (2003) and other papers that did find a significant win and loss effect.
5. Conclusion
Motivated by the abundance of psychological evidence indicating that sports sentiment has a significant impact on investor’s mood, this study investigates the possible relationship between stock returns on national stock markets and football results of nation teams. By including new data from 1999 up to and including 2018, this study has contributed to the existing debate on sports sentiment. Contrary to previous studies, this study controls for the ability bias by using a surprise variable that is not simply a dummy variable. This variable can take higher and lower values to distinguish between different level of surprises. A GARCH(1,1) model is included to take into account the time varying volatility of stock returns. It appears that there is neither a significant relationship between positive returns on national stock market and football wins nor a significant relationship between negative returns on those markets and football losses. Several robustness checks do not improve the significance of the results. The null hypothesis that football results do not affect stock markets cannot be rejected. There are no deviations from the efficient market hypothesis observed that are caused by changes in investor’s behaviour after football games. Therefore, the efficient market hypothesis is supported. Investors appear to act rational when it comes to football sentiment. It is hard for investors to take advantage of arbitrage opportunities after its nation has played at the World Cup or at any of the continental cups.
25 the paper of Fan and Wang (2018), that controlled for small-cap stocks. Thirdly, previous studies on sports sentiment often state that qualifying games are of less importance than games during the specific tournament itself. Therefore, qualifying games are often neglected. However, it can be argued that qualifying games are as important as tournament games. This can be especially the case for countries that are close to qualify for the World Cup or any continental cup. This fact needs to be taken into account more carefully. Perhaps there is a significant win and loss effect present for qualification games, while there is no significant effect for games during the tournament. Lastly, there is a lack of research done in the field of sports sentiment for other sports. Edmans et al. (2007) included cricket, basketball and rugby, but there are many more sports that might have an effect on investors’ sentiment. To conclude, new studies in the field of sports sentiment need to carefully take into account the words of Klein et al. (2009) that significant relationships can be a sheer coincidence. Studies can suffer from publication bias that enlarge datasets and include many variables to find more significant results. It is important that new studies critically evaluate previous studies to contribute to the still ongoing debate on sports sentiment and stock returns.
26 References
Articles
Ashton, J.K., Gerrard, B. and Hudson, R. (2003), ‘Economic impact of national sporting success: evidence from the London stock exchange’, Applied Economics Letters, Volume 10, 783-785
Ashton, J.K., Gerrard, B. and Hudson, R. (2011), ‘Do national soccer results really impact on the stock market?’, Applied Economics, Volume 43, 3709-3717
Berument, M.H., Ceylan, N.B. and Ogut-Eker, G. (2009), ‘Soccer, stock returns and fanaticism: Evidence from Turkey’, The Social Science Journal, Volume 46, 594-600 Berument, M.H. and Ceylan, N.B. (2012), ‘Effects of soccer on stock markets: The return-volatility relationship’, The Social Science Journal, Volume 49, 368-374
Berument, M.H. and Ceylan, N.B. (2013), ‘Soccer and stock market risk: empirical evidence from the Istanbul Stock Exchange’, Psychological Reports: Sociocultural Issues in
Psychology, Volume 3, 763-770
Bollerslev, T. (1986), ‘Generalized autoregressive conditional heteroskedasticity’, Journal of
Econometrics, Volume 31 (3), 307–327.
Copeland, M. and Copeland, T. (1998), ‘Leads, Lags, and Trading in Global Markets’,
Financial Analysts Journal, Volume 54 (4), 70-80
Cross, F. (1973), ‘The Behavior of Stock Prices on Fridays and Mondays’, Financial Analysts
Journal, Volume 29, 67-69
Curatola, G., Donadelli, M., Kizys, R. and Riedel, M. (2016), ‘Investor Sentiment and Sectoral Stock Returns: Evidence from World Cup Games’, Finance Research Letters, Volume 17, 267-274
Dahlquist, M. and Robbertson, G. (2001), ‘Direct foreign ownership, institutional investors, and firm characteristics’, Journal of Financial Economics, Volume 59, 413-440
27 Ehrmann, M. and Jansen, D. (2016), ‘It Hurts (Stock Prices) When Your Team is about to Lose a Soccer Match’, Review of Finance, 1215-1233
Edmans, A., García, D. and Norli, Ø (2007), ‘Sports Sentiment and Stock Returns’, The
Journal of Finance, Volume 62 (4), 1967-1998
Fama, E. F. (1970), ‘Efficient Capital Markets: A Review of Theory and Empirical Work’,
The Journal of Finance, Volume 25 (2), 383-417
Fan, Q. and Wang, T. (2018), ‘Game day effect on stock market: Evidence from four major sports leagues in US’, Journal of Behavioral and Experimental Finance, Volume 20, 9-18 Floros, C. (2014), ‘Football and Stock Returns: New Evidence’, Procedia Economics and
Finance, Volume 14, 201-209
Frieder, L. and Subrahmanyam, A. (2004), ‘Nonsecular regularities in returns and volume’,
Financial Analysts Journal, Volume 60
Geyer-Klingeberg, J., Hang, M., Walter, M. and Rathgeber, A. (2018), ‘Do stock markets react to soccer games? A meta-regression analysis’, Applied Economics, Volume 50 (19), 2171-2189
Hirshleifer, D. and Shumway, T. (2003), ‘Good day sunshine: Stock returns and the weather’,
Journal of Finance, Volume 58, 1009-1032
Kahneman, D. and Tversky, A. (1979), ‘Prospect theory: An analysis of decision under risk’,
Econometrica, Volume 47, 263-292
Kamstra, M., Kramer, L. and Levi, M. (2000), ‘Losing sleep at the market: The daylight saving anomaly’, American Economic Review, Volume 12, 1000-1005
Kamstra, M., Kramer, L. and Levi, M. (2003), ‘Winter blues: A SAD stock market cycle’,
American Economic Review, Volume 93, 324-343
Kaplanski, G. and Levy, H. (2010), ‘Exploitable Predictable Irrationality: The FIFA World Cup Effect on the U.S. Stock Market’, The Journal of Financial and Quantitative Analysis, Volume 45 (2), 535-553
Klein, C., Zwergel, B. and Heiden, S. (2009), ‘On the existence of sport sentiment: the relation between football match results and stock index returns in Europe’, Review of
28 Kloner, R., McDonald, S., Leeka, J. and Poole K. (2009), ‘Comparison of Total and
Cardiovascular Death Rates in the Same City During a Losing Versus Winning Super Bowl Championship’, The American Journal of Cardiology, Volume 103 (12), 1647-1650
Koning, R.H. and McHale, I.G. (2012), ‘Estimating match and World Cup winning probabilities’. In W. Maennig, & A. Zimbalist (Eds.), International handbook on the
economics of mega sports events, 177-193. London: Edward Elgar Publishing
Markman, K. and Hirt, E. (2002), ‘Social prediction and the ‘allegiance bias’’, Social
Cognition, Volume 20, 58-86
Pollard, R. (2008), ‘Home advantage in football: A current review of an unsolved puzzle’,
The Open Sports Sciences Journal, Volume 1, 12-14
Pollard, R. and Gómez M. (2014), ‘Components of home advantage in 157 national soccer leagues worldwide’, International Journal of Sport and Exercise Psychology, Volume 12 (3), 218-233
Wann, D., Dolan, T., McGeorge, K. and Allison, J. (1994), ‘Relationships between spectator identification and spectators' perceptions of influence, spectators' emotions, and competition outcome’, Journal of Sport and Exercise Psychology, Volume 16 (4), 347-364
Websites
Asian Football Confederation (2019), Stats AFC, accessed March 3 2019, <stats.the-afc.com> Confederación Sudamericana de Fútbol (2019), Copa América, accessed March 3 2019, <https://copaamerica.com/es>
Confederation of African Football (2019), Total Africa Cup of Nations, accessed March 3 2019, <www.cafonline.com/en-us/home.aspx>
Confederation of North, Central American and Caribbean Association Football (2019), Gold
Cup, accessed March 3 2019, <https://www.goldcup.org>
29 Féderation Internationale de Football Association (2018), FIFA Tournaments – Archive, accessed March 3 2019, <https://www.fifa.com/fifa-tournaments/archive>
Fitzgerald, T. (2018), FIFA Soccer Rankings, Kaggle, accessed April 19 2019,
<https://www.kaggle.com/tadhgfitzgerald/fifa-international-soccer-mens-ranking-1993now> Jürisoo, M. (2018), International Football Results from 1872 to 2018, Kaggle, accessed 22 February 2019, <https://www.kaggle.com/martj42/international-football-results-from-1872-to-2017>
Klapper, L., Lusardi, A. and Van Oudheusden, P. (2015), Financial Literacy Around the
World, World Bank Development Research Group, accessed 3 April 2019,
<https://responsiblefinanceforum.org/wp-content/uploads/2015/12/2015-Finlit_paper_17_F3_SINGLES.pdf>
Publicis Media Sport & Entertainment (2018), 2018 FIFA World Cup Russia™: Global Broadcast and Audience Summary, accessed 16 March 2019,
<https://resources.fifa.com/image/upload/2018-fifa-world-cup-russia-global-broadcast-and-audience-executive-summary.pdf?cloudid=njqsntrvdvqv8ho1dag5>
Union of European Football Associations (2019), UEFA EURO – History, accessed March 3 2019, <https://www.uefa.com/uefaeuro/history/index.html>
World Atlas (2018), The Most Popular Sports, accessed 5 March 2019, <https://www.worldatlas.com>
30 Appendix
Table A1: Stock market index per country
Table A2: Non-weekend (public) holidays per country
Country All non-weekend holidays5
Argentina New Year’s Day, Malvinas War Veterans Day, Jueves Santo, Viernes Santo, Labour Day, Revolution Day, General Martín Manuel de Güemes Memorial Day, General Belgrano Memorial Day, Independence Day, Day of the
Immaculate Conception, Christmas Day, San Martín Memorial Day, Columbus Day, National Sovereignty Day
Australia New Year’s Day, Australia Day, Good Friday, Easter Monday, Anzac Day, Christmas Day, Boxing Day
Brazil New Year’s Day, Good Friday, Tiradentes Day, Labor Day, Corpus Christi, Independence Day, Day of Nossa Senhora de Aparecida, All Souls’ Day, Proclamation of the Republic, Christmas Day
Canada New Year’s Day, Good Friday, Easter Monday, Victoria Day, Canada Day, Labour Day, Thanksgiving, Remembrance Day, Christmas Day, Boxing Day Chile New Year’s Day, Good Friday, Labour Day, Navy Day, Feast of Saints Peter and Paul, Assumption, National Unity Day, Independence Day, Army Day,
5 All data on the non-weekend holidays is gathered from World Travel Guide
Country Stock market index
Argentina MERVAL
Australia ASX 200
Brazil Bovespa
Canada TSX Composite Index
Chile IPSA
Colombia1 IGBC
Costa Rica IACR
Ecuador Quayaguil Select
Egypt EGX 30
England FTSE 100
France CAC 40
Germany DAX 30
Italy FTSE MIB
Japan Nikkei 225
Korea Republic Kospi
Mexico IPC
Netherlands AEX
Nigeria NGSE All Share
Peru Lima General
Portugal PSI 20
Saudi Arabia Tadawul South Africa JSE All Share
Spain IBEX 35
Tunisia Tunis Main Index
31 Discovery of Two Worlds Day, All Saint’s Day, Immaculate Conception,
Christmas Day
Colombia New Year’s Day, Epiphany, St Joseph’s Day, Maundy Thursday, Good Friday, Labour Day, Ascension Day, Corpus Christi, Independence Day, Battle of Boyac, Assumption, Columbus Day, All Saint’s Day, Independence of Cartagena City, Immaculate Conception, Christmas Day
Costa Rica New Year’s Day, Juan Santamar, Labour Day, Guanacaste Day, Feast of Patroness, Assumption, Independence Day, Columbus Day, Christmas Day Ecuador New Year’s Day, Good Friday, Labour Day, Battle of Pichincha, Independence
Day, Guayaquil Independence Day, All Soul’s Day, Cuenca Independence Day, Christmas Day, New Year’s Eve
Egypt Sinai Liberation Day, Labour Day, National Day, Armed Forces Day England New Year’s Day, Good Friday, Easter Monday, Early May Bank Holiday,
Spring Bank Holiday, Summer Bank Holiday, Christmas Day, Boxing Day France New Year’s Day, Easter Monday, Labour Day, Victory Day 1945, Ascension
Day, Whit Monday, Bastille Day, Assumption, All Saint’s Day, Armistice Day, Christmas Day
Germany New Year’s Day, Epiphany, International Women’s Day, Good Friday, Easter Monday, Labour Day, Ascension Day, Whit Monday, Day of German Unity, Day of Reformation, All Saint’s Day, Repentance Day, Christmas Day Italy New Year’s Day, Epiphany, Easter Monday, Liberation Day, Labour Day,
Republic Day, Assumption, All Saint’s Day, Immaculate Conception, Christmas Day, St Stephen’s Day
Japan New Year’s Day, Coming of Age Day, National Foundation Day, Showa Day, Constitution Memorial Day, Greenery Day, Children’s Day, Marine Day, Respect for the Aged Day, Health and Sports Day, Culture Day, Labour Thanksgiving Day, Emperor’s Birthday
Korea Republic New Year’s Day, Sam Il Jul, Labour Day, Orininal, Memorial Day, Kwang Bok Jul, Kae Chun Jul, Christmas Day
Mexico New Year’s Day, Constitution Day, Birthday of Benito Ju, Maundy Thursday, Good Friday, Labour Day, Anniversary of the Battle of Puebla, Independence Day, Day of the Race, Día de Muertos, Revolution Day, Day of Our Lady of Guadalupe, Christmas Day
Netherlands New Year’s Day, Good Friday, Easter Monday, King’s Day, Liberation Day, Ascension Day, Whit Monday, Christmas Day, Boxing Day
Nigeria New Year’s Day, Good Friday, Easter Monday, Workers’ Day, Independence Day, Christmas Day, Boxing Day
Peru New Year’s Day, Maundy Thursday, Good Friday, Labour Day, Feast of Saints Peter and Paul, Independence Day Celebrations, St Rosa of Lima Day, Battle of Angamos, All Saint’s Day, Immaculate Conception, Christmas Eve, Christmas Day
Portugal New Year’s Day, Good Friday, Easter Sunday, Freedom Day, Labour Day, Portugal Day, Corpus Christi, Assumption, Republic Day, All Saint’s Day, Restoration of Independence Day, Immaculate Conception, Christmas Day Saudi Arabia National Day
South Africa New Year’s Day, Human Rights Day, Good Friday, Family Day, Freedom Day, Workers’ Day, Youth Day, National Women’s Day, Heritage Day, Day of Reconciliation, Christmas Day, Day of Goodwill
Spain New Year’s Day, Epiphany, Day of the Balearic Islands, St Joseph’s Day, Maundy Thursday, Good Friday, Labour Day, Assumption, National Day, All Saint’s Day, Constitution Day, Immaculate Conception, Christmas Day, Boxing Day
32 USA New Year’s Day, Martin Luther King, Jr. Day, Presidents’ Day, Memorial Day,
Independence Day, Labour Day, Columbus Day, Veterans’ Day, Thanksgiving Day, Christmas Day
Table A3: Average FIFA World Ranking per country (1999-2018)
Figures A1-A25: Daily return on national stock market index per country (1999-2018)
Figure A1: MERVAL (Argentina) Figure A2: ASX 200 (Australia)
6 Top 13 countries (from Spain up to and including USA) are ‘football giants’. Countries ranked 14-25 (from Ecuador up to and including Canada) are ‘football dwarfs’. Rankings are based on average points.
Ranking6 Country Average points Average ranking
33
Figure A3: Bovespa (Brazil) Figure A4: TSX Composite Index (Canada)
Figure A5: IPSA (Chile) Figure A6: IGBC (Colombia)
Figure A7: IACR (Costa Rica) Figure A8: Quayaguil Select (Ecuador)
Figure A9: EGX 30 (Egypt) Figure A10: FTSE100 (England)
34
Figure A11: CAC 40 (France) Figure A12: DAX 30 (Germany)
Figure A13: FTSE MIB (Italy) Figure A14: Nikkei 225 (Japan)
Figure A15: Kospi (Korea Republic) Figure A16: IPC (Mexico)
Figure A17: AEX (Netherlands) Figure A18: NGSE All Share (Nigeria)
35
Figure A19: Lima General (Peru) Figure A20: PSI 20 (Portugal)
Figure A21: Tadawul (Saudi Arabia) Figure A22: JSE All Share (South Africa)
Figure A23: IBEX 35 (Spain) Figure A24: Tunis Main Index (Tunisia)
Figure A25: S&P 500 (USA)
