1
The Effects of Football Game
Outcomes on Stock Price
Performance
Which football game variables affect a club’s stock price?
Abstract
To what extent do football game outcomes affect a club’s stock prices? This research investigates the investors' reaction to football results of 20 soccer clubs in Europe over the period 2013-2017. The study controls for several variables and one of these variables is the expected outcome of a soccer game, which is determined based on bookmakers’ betting odds. The main findings of this research are that wins generate a significantly positive effect on a club’s stock price, whereas draws and losses significantly affect the stock prices negatively on the first trading day after a game. The higher the surprise of a win, the higher are the club’s abnormal returns. Moreover, both a large goal difference and a small club size generate a stronger stock price reaction than a small goal difference and large club size.
Student number: S2384523 Name: Anna Koerhuis
Study Program: MSc Finance
Supervisor: Prof. Dr. Wolfgang Bessler*
Field Keywords: Rational effects; Sentiment effect; Event Study; Stock returns; Football results; football clubs
Words: 12,940
2 1. Introduction
In the past decades, several football clubs decided to go public to meet their financial needs. The valuation of stocks from listed football clubs is an interesting subject, as it not only depends on financial performance but also on sports performance. The current study investigates the stock price reaction following after different football game outcomes. Usually, wins (positive outcomes) should result in positive stock price reactions, whereas for losses (adverse outcomes) the opposite may hold (see, e.g. Palomino, Renneboog, and Zhang, 2009). The size of the stock price reaction may also depend on the deviation from the expected outcome, i.e. when an underdog team wins a game.
This research investigates the following primary question: ‘Which variables from football games can explain clubs' stock price reactions?' The study includes many variables related to football games that can influence stock prices. These variables include real game outcome, expected game outcome, goal difference, venue, importance and size of the football club. Previous literature shows two general explanations of stock price reactions. First, the rational effect includes the economic effects of football games, such as higher television rights revenues, ticket sales or sponsorship revenues (see, e.g. Palomino and Sákovics, 2004). Second, the sentiment effect covers price reactions from investors’ irrational behavior (see, e.g. Ashton, Gerrard, and Hudson, 2003). Hence, it is important to analyze the investors’ reaction to various types of football results by controlling for the expected outcome. This research uses betting odds as a proxy for the prediction of football game outcomes. According to several studies, e.g. from Forrest, Goddard, and Simmons (2005), odds from betting markets should provide the most reliable and unbiased expectations of the outcome of a football game.
The main findings show that won and lost football games result in respectively positive and negative stock price reactions, where the loss effect is substantially larger than the win effect. The findings show ambiguous results of the test whether unexpected results cause stronger stock market responses than expected results. However, when game outcomes cause a stock reaction, this reaction is directly reflected in the stock prices on the first trading day after a game1. Besides, the results from the analysis show that games with emotional load and games of higher importance generate higher abnormal returns.
Previous studies investigate a similar effect for one market, or for one or two variables that can explain stock price reactions after games (see, e.g. Palomino, Renneboog, and Zhang, 2009). This research contributes to the existing literature by investigating within a sample of 2,482 football games, a large number of variables that can explain abnormal stock prices. Moreover, the current study supports the main findings of the study from Scholtens and Peenstra (2009) and the recent study from Dimic et al. (2018). It is interesting to investigate the
1 The event date itself is not included in the analysis, because football games often take place in weekends or at a
3 primary variable that explains stock price reactions since investors can respond to future expected price changes and profit from it. The sample of this study embraces 20 publicly traded football clubs in Europe. The sample period includes the four seasons from 2013-14 to 2016-17. Selection criteria for the sample are that the club is listed during the whole sample period, and that the betting odds for each game are available. The research used an event study methodology to determine whether returns on the first (and sequential) trading days are significantly different from expected returns. The market model measures the expected returns, with a country’s market index as a benchmark.
4 2. Literature review
This section discusses the existing literature with respect to the effect of football match outcomes on the football club’s stock prices. Several studies perform investigations comparable to the current study but for a different sample of leagues, clubs, countries, sample period or sample size. Section 2.1 discusses two competing explanations for abnormal returns: rational expectations and sentiment effects. Section 2.2 reports previous studies on the effect of football game outcomes on the stock market, on which the hypotheses in the current study are based. Section 2.4 discusses different ways to predict football game outcomes and argues which prediction method performs best.
2.1 Rational and sentiment effects
According to Palomino, Renneboog, and Zhang (2009), football clubs’ stock price increases (decreases) from won (lost) games can be at least partly explained by rational arguments. Falconieri, Palomino, and Sákovics (2004) and Palomino and Sákovics (2004) address the economic effects of football games. First, revenues from the national television deals are performance-based distributed to the participating teams, i.e. based on the ranking at the end of the season. Second, moving from a lower to a higher league results in an increase in the income from television rights. Third, higher performance generally results in a higher ticket price and increased merchandising and sponsorship revenues. The subsequent studies investigate performance-based economic effects of football outcomes. Bernile and Lyandres (2011) argue that winning the European Champions League or UEFA Cup positively affects the club’s return on assets. Barajas, Fernández-Jardón, and Crolley (2005) found that Spanish football clubs gain higher club revenues when they are performing well. Berument and Yucel (2005) found that the growth rate of Turkish industrial output increases when Fenerbahce won a match in the European competition. Moreover, Berument, Ceylan, and Gozpinar (2006) found an increase in the Turkish stock market after Besiktas won a match.
5 visible on the US market. The outcomes of the study show that a win of a team can cause changes in behavior: people are more confident, more optimistic, and less risk-averse.
2.2 Construction of the hypotheses
The first systematic study on the effect of football matches on the club’s stock price is reported by Morrow (1999), who has analyzed games of Manchester United and Sunderland in the UK. His results indicate a negative effect after losses and a positive effect after wins, while controlling for stock market-wide price changes. Renneboog and Van Brabant (2000) investigate stock prices of listed British football clubs and find a significant positive reaction after wins, and a negative reaction after draws and losses in the national cup and European competition. Each subsequent paragraph discusses different subjects within the effect of football game outcomes on stock prices that are reported in previous studies. From the findings of these previous studies, the hypotheses for the current study will be derived.
The first hypothesis involves the general effect of football game outcomes. Edmans, García, and Norli (2007) found a significant negative national stock market reaction after losses in the continental cups and World Cup and no general reaction after wins. Scholtens and Peenstra (2009) found significant positive stock price reactions after wins and negative reactions after losses in national and European football leagues, measured from 2000 to 2004. In the same vein, Dimic et al. (2018) examine stock price reactions of league matches from a sample of thirteen publicly traded football clubs that are located in six countries. Their study shows positive significant abnormal returns on the first trading day after a won game. Besides, the stock prices respond negatively after a loss, which effect is larger than the effect after a win. Other studies confirm the negative effect after losses and positive effect after wins for the subsequent markets: Germany (Stadtmann, 2006), Turkey (Saraç and Zeren, 2013), Portugal (Duque and Ferreira., 2005) and The United Kingdom (Benkaiem, Le Roy and Louhichi, 2011). Findings of the previously mentioned studies lay the foundations of the first hypothesis: Positive football game outcomes result in positive abnormal stock returns of the club, while negative outcomes generate negative abnormal returns.
6 The different effect of expected and unexpected game outcomes are included in the third hypothesis. This paragraph addresses studies that investigate stock price effects of predicted and unpredicted football outcomes, which show contrasting results. According to the efficient market hypothesis, publicly available information is reflected in the stock prices. Therefore, it is natural to assume that reliable predictions of football games are reflected in the stock prices before the real outcome is known. Findings from the study of Palomino, Renneboog, and Zhang (2009) support this hypothesis with respect to losses: results show a higher stock market response after unexpected losses than after expected losses. Conversely, Scholtens and Peenstra (2009) show that for national leagues, expected losses cause a significantly stronger response on the stock market than unexpected losses. Dimic et al. (2018) show a negative and vast stock price reaction after losses, regardless of whether the result was expected or not, which is the so-called surprise level. Focusing on the wins, Scholtens and Peenstra (2009) suggest that expected wins do not have a significant effect on stock prices, while unexpected wins have a significant positive effect on stock prices. Dimic et al. (2018) report that the price reaction after wins increases with the surprise level. Palomino, Renneboog and Zhang (2009) found distinct results that show higher abnormal returns after strongly expected wins than weakly expected wins. The results of these studies suggest that the publicly available information on betting quotes is not fully absorbed by the market, which implies irrational stock market behavior and reactions driven by investors' sentiment. The third hypothesis tests the presence of the surprise effect or sentiment effect, which is the following: unexpected game outcomes cause a stronger stock price response than expected outcomes.
7 The presence of emotional load in a game will be addressed in the fifth hypothesis. Events that have a higher emotional load, such as a large goal difference or the location of the game, are resulting in stronger abnormal price reactions (Dimic et al., 2018). The studies of Benkraiem, Louhichi, and Marques (2009) and Benkraiem, Le Roy and Louhichi (2011) investigate the role of the venue of the football game for European listed clubs. These studies show that home losses a stronger price reaction than losses away. Losing a game at home can give implications for future games: since the club cannot take advantage of the home ground, it is questionable whether they can win in the future. These findings provide suggestions for the fifth hypothesis that will be tested: Emotional load in a game causes higher abnormal returns, including the goal difference and the venue of the game.
The sixth hypothesis entails the importance level of a football game, which can influence the magnitude of the stock price reaction. Renneboog and Van Brabant (2000) report that relegation or promotion games generates substantial abnormal returns for British football clubs. This substantial effect can be economically explained since the Premier League and European games assure much higher (future) income with respect to sponsoring income and television broadcasting rights. Edmans, García, and Norli (2007) show in continental cups and the World Cup, the loss effect on the national stock market for elimination games is larger than for group games. Bell et al. (2012) perform a similar examination and address two determinants for the importance of the match. First, a match with two rivals, of which the two teams are competing for the same positions. Second, when a team is likely to promote or relegate at the end of the season. The results show higher affected stock prices for a match of high importance than a match of less importance. However, despite these significant results, effects from game outcomes appear to be modest compared to changes in a club’s stock prices caused by other variables. The sixth hypothesis tests whether games of higher importance generate significant stronger stock reactions than games of less importance.
The seventh and last hypothesis focuses on the size of football clubs. Smaller stocks are more strongly affected by investor sentiment than larger stocks (Baker and Wurgler, 2006). Edmans, García, and Norli (2007) report that that small stocks generally have a higher local ownership and are more sensitive than large stocks. Their findings show that smaller football clubs experience stronger stock responses from game results than large football clubs2. Therefore, the seventh hypothesis that will be studied in the current research is that smaller football clubs experience a stronger stock reaction from games than large football clubs do.
2 The index selection for clustering the clubs based on size is described in the appendix of Edmans, García, and
8 2.3 The prediction of football outcomes
Several studies investigate methods to predict football match outcomes. Existing literature describes several statistical models to forecast football outcomes and test the performance of these models against real outcomes, other models and betting odds. This section describes the outcomes of these studies.
9 3. Research method
Chapter 3 describes the methods used to conduct the current event study. Section 3.1 describes the approach of estimating the expected and abnormal returns on the first trading day after the match. Section 3.2 illustrates the approach to derive expectations from betting odds. To analyze the effect of football game results on stock market returns, the event study method is well suited due to the subsequent arguments.
• New information is frequently available.
• The event outcome is unknown in advance, which implies that the matches are not settled in advance so that no pre-event reaction exist.
• The event outcome can be quantified (number of goals, received points after a match).
• The time that new information arrives is precise and is immediately publicly available when the match ends.
• Betting odds show the expectations of the matches and therefore quantify market expectations.
3.1 The estimation of returns
In the current event study, the performance of the stocks is measured by the market model, which is linear regressed. The market model describes the return of a given security by the return of the market portfolio (MacKinlay, 1997). The assumed joint normality of security returns results in a linear one-factor model. For security i, the return, R, at time t is calculated as
𝑅𝑅𝑖𝑖𝑖𝑖 = 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑖𝑖𝑅𝑅𝑚𝑚𝑖𝑖+ 𝜀𝜀𝑖𝑖𝑖𝑖 𝐸𝐸(𝜀𝜀𝑖𝑖𝑖𝑖 = 0) 𝑣𝑣𝑣𝑣𝑣𝑣(𝜀𝜀𝑖𝑖𝑖𝑖) = 𝜎𝜎𝜀𝜀𝑖𝑖2
where 𝑅𝑅𝑖𝑖𝑖𝑖 and 𝑅𝑅𝑚𝑚𝑖𝑖 represent respectively the returns on security and the market portfolio for period t. The error term 𝜀𝜀𝑖𝑖𝑖𝑖 is the difference between the actual and predicted return that has a zero mean. 𝛼𝛼𝑖𝑖, 𝛽𝛽𝑖𝑖, and 𝜎𝜎𝜀𝜀2 represent the parameters of the market model. Adding the market component to the basic constant mean return model decreases the variance of the abnormal return. This addition, in turn, can increase the capability of identifying event effects. The second reason to prefer the market model over the constant mean return model is the use of an estimation window. Due to the high frequency of football matches, the estimation windows and event windows will overlap when estimation windows are larger than three days. When the event windows are overlapping and if the covariance between the abnormal returns is not equal to zero, generated results of aggregated abnormal returns are no longer valid. Besides, the market model regression provides an R2, which shows the benefit of using the model. A higher R2 results in a lower abnormal return variance, which again increases the capability of identifying event effects and thus increases the benefit of the model. Moreover, this study applies the approach of Brown and Hartzell (2001) of using the whole sample period as
10 estimation period, which is also used by e.g. Scholtens and Peenstra (2009) as a robustness check. In particular, the approach processes all return data to estimate normal returns. The following regression approach measures the abnormal return, 𝐴𝐴𝑅𝑅𝑖𝑖𝑖𝑖, after a match by
𝐴𝐴𝑅𝑅𝑖𝑖𝑖𝑖 = 𝑣𝑣𝑖𝑖𝑖𝑖− 𝛼𝛼�𝑖𝑖− 𝛽𝛽̂𝑖𝑖𝑣𝑣𝑚𝑚𝑖𝑖 ,
where 𝑣𝑣𝑖𝑖𝑖𝑖 is the observed return and 𝛼𝛼�𝑖𝑖 and 𝛽𝛽� are respectively the estimators for the parameters 𝚤𝚤 𝛼𝛼𝑖𝑖 and 𝛽𝛽𝑖𝑖. Further, the general statistical hypothesis that is tested is the following:
𝐻𝐻0: 𝐴𝐴𝐴𝐴𝑅𝑅 = 0 𝐻𝐻1: 𝐴𝐴𝐴𝐴𝑅𝑅 ≠ 0
If the average abnormal return (AAR) is not significantly different from zero, the model can predict the return sufficiently. An AAR that significantly differs from zero shows that these events significantly differ from the usual situation. The Student t-test evaluates the significance of the AAR. The cumulative average abnormal Return (CAAR) is the AAR over subsequent i days after the event and can be specified by
𝐶𝐶𝐴𝐴𝐴𝐴𝑅𝑅𝑖𝑖 = ∑ 𝐴𝐴𝑅𝑅𝑖𝑖
The described procedure is carried out for the whole sample of events in different compositions, based on wins, losses, or draws, expected outcome, goal difference, country, club size, and home or away games. A club’s stocks can have a different sensitivity to these different aspects.
3.2 Returns from betting odds
As Section 2.3 addresses, betting odds is a reliable proxy to derive expected football game results. This section shows the conversion of betting odds for specific football games into probabilities of outcomes, which is a similar method to the ones used by Palomino, Renneboog, and Zhang (2009) and Dimic et al. (2018). The following formulas reflect the bookmakers' beliefs and represent the normalized probabilities to win and to lose, respectively.
𝑃𝑃𝑣𝑣𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖 = 𝑥𝑥𝑖𝑖𝑖𝑖 −1 𝑥𝑥𝑖𝑖𝑖𝑖−1+ 𝑥𝑥 𝑖𝑖𝑖𝑖−1+ 𝑥𝑥𝑖𝑖𝑖𝑖−1 𝑃𝑃𝑣𝑣𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖 =𝑥𝑥 𝑥𝑥𝑖𝑖𝑖𝑖−1 𝑖𝑖𝑖𝑖−1+ 𝑥𝑥𝑖𝑖𝑖𝑖−1+ 𝑥𝑥𝑖𝑖𝑖𝑖−1
𝑋𝑋𝑖𝑖𝑖𝑖 expresses one plus the betting odds for a bet on the match outcome, where 𝑃𝑃 represents the team and 𝑗𝑗 the outcome of the match (win, draw or loss). Where 𝑤𝑤, 𝑑𝑑, and 𝑙𝑙 expresses a win, draw, and loss, respectively. For a match of team 𝑃𝑃 with outcome 𝑗𝑗, the bettor receives 𝑥𝑥𝑖𝑖𝑖𝑖 units of money for each unit of money bet. The probability of a draw is indirectly captured in the preceding formulas, when both formulas generate a small probability of a win and loss.
A second measure represents the game outcome uncertainty, which is also determined by bookmakers’ beliefs on the outcome. Probdiffi shows the difference between the probability
to win and to lose a match of team 𝑃𝑃 by
11 A positive (negative) value of Probdiffi represents a higher probability of winning (lose) than
to lose (win). The closer the absolute value of Probdiffi is to zero, the higher the uncertainty of
the game outcome.
The generated values of the probabilities of a win or loss and the difference can be grouped in strongly expected to win, weakly expected to win, weakly expected to lose, and strongly expected to lose. Based on Palomino et al. (2009), dummy variables indicate the specific group to which the match attributes and are measured in two manners. Measure [a] is determined by ProbWin and measure [b] is determined by Probdiffi:
• SEW (strongly expected to win): When ProbWini > 0.45, SEW[a] corresponds to one, and alternatively zero. When Probdiffi > 0.3, SEW[b] corresponds to one and
alternatively zero.
• WEW (weakly expected to win): When ProbWini = [0.35,0.45], WEW[a] corresponds to one, and alternatively zero. When Probdiffi = [0,0.3], WEW[b] corresponds to one
and alternatively zero.
• WEL (weakly expected to lose): When ProbWini = [0.25,0.35], WEL[a] corresponds to one, and alternatively zero. When Probdiffi = [-0.3,0], WEL[b] corresponds to one and
alternatively zero.
• SEL (strongly expected to lose): When ProbWini < 0,25, SEL[a] corresponds to one, and alternatively zero. When Probdiffi < -0.3, SEL[b] corresponds to one and
12 4. Data
This research includes 20 listed European football clubs over four seasons: 2013-14 to 2016-17. The sample embraces all games from the highest national league of the subsequent ten countries: Denmark (Superliga), Netherlands (Eredivisie), France (Ligue 1), Germany (Bundesliga), Italy (Serie A), Poland (Ekstraklasa), Portugal (Primeira Liga), Scotland (Premiership), Sweden (Allsvenskan), and Turkey (Superliga). Table 1 presents the included clubs in the sample, their average market value over the period 2013-2017 and their final standing for the seasons 2013-14 to 2016-17. The market value of the clubs varies between 1 and 323 million Euros. Daily stock returns and balance sheet and income statement items are retrieved from Thomson Financial Datastream. Table 10 in Appendix A shows more detailed information of the clubs included in the sample, such as country and IPO date.
Table 1. Descriptive details of listed football clubs included in the current study’s sample. MV Standing national league
Club € Million 2013-14 2014-15 2015-16 2016-17 Aalborg 9.05 1 5 5 8 Aarhus 8.82 11 14* 10 12 AIK 3.57 3 3 2 2 Ajax 161.87 1 2 2 2 AS Roma 173.67 2 2 3 2 Benfica 24.18 1 1 1 1 Besiktas 213.52 3 3 1 1 Brondby 24.36 4 3 4 2 Celtic 88.91 1 1 1 1 Dortmund 368.57 2 7 2 3 FC Copenhagen 93.48 2 2 1 1 FC Porto 9.90 3 2 3 2 Fenerbahce 305.08 1 2 2 3 Galatasaray 162.64 2 1 6 4 Juventus 323.02 1 1 1 1 Lazio 38.43 9 3 8 5 Lyon 85.50 5 2 2 4 Ruch 1.41 3 14 8 14 Silkeborg 9.83 13* 12 14* 9 Sporting 23.98 2 3 2 3
This table presents the subsequent details of the football clubs included in the sample of this study: average market value (MV) in million euros from the period 2013-2017 and the standing in the country’s highest national league during the sample period of 2013-14 to 2016-17.
13 Data selection criteria are that clubs need to be listed during the whole sample period, and all football games must have a betting quotation. Oddsportal provides data on betting quotes and match results. This study treats the betting odds of up to 11 bookmakers as a general expectation from football specialists. These betting odds show how much money one earns when winning the bet. The sample excludes results from non-listed clubs. Moreover, to prevent that the same game is included twice, for games in which the opponents are both listed, only one club will be observed. In these cases, the club that is included in the sample and thus will be observed, is randomly selected. After these adjustments, the final sample consists of 2,482 match observations. Appendix A presents more detailed information about the game outcomes per country in Table 11 and about the benchmarks in Table 12. Country indices are used as a benchmark each market. Moreover, the STOXX Europe Football Index serves as a benchmark for the European football clubs. This index covers all European football clubs listed on a stock exchange.
Table 2a shows the number and percentage of wins, draws and losses for the different venues. Games played at home are won 68% of the time, while games away are won 51% of the time. This difference suggests the presence of a home ground advantage. From the club’s details presented in Table 2b, one can conclude that the average standing at the end of the season is substantially lower for small clubs than for medium and large clubs. This relation could be economically explained, since clubs that have a high market value usually have a higher budget for high quality football players or trainers. These players and trainers can play a large role in attaining a higher performance level and thus a higher standing.
Table 2a. Outcome details based on venue. Table 2b. Club’s details per size group.
Win Draw Loss
Avg. Size (€ MM) No. Of clubs Avg. Standing Total 1469 526 487 59% 21% 20% Small 7.1 6 7.21 Home 841 224 180 Medium 54.1 7 2.68 68% 18% 14% Large 244.1 7 2.25
Away 628 302 307 This table shows summarized details of listed football clubs that are included in the sample per size group. The average standing is measured over the period 2013-2017
51% 24% 25%
This table shows the number and distribution of football game results that are included in this study's sample.
14 events, alternative [a] and [b] generate a similar expectation. However, for the events that do not have a clear expectation, the superior alternative should provide this expectation. By comparing the number of proper forecasts, alternative [b] appears to provide a better forecast of the game outcome than alternative [a] (see Appendix C for the number and distribution of expected and real football game outcomes, alternative [a] over [b]). Evidently, listed football clubs are on average performing quite well, since the number of wins are significantly higher than the number of draws and losses. To illustrate, 59% of the included games are won, compared to 21% losses and 20% draws. All observations are football games from a country’s highest league.
Table 3. The number and distribution of expected and real football game outcomes.
Expected outcome Real outcome Total
Wins Draws Losses games
Strongly expected to win 1056 257 154 1467
72% 18% 10% 100%
Weakly expected to win 299 136 152 587
51% 23% 26% 100%
Weakly expected to lose 100 109 119 328
30% 33% 36% 100%
Strongly expected to lose 14 24 62 100
14% 24% 62% 100%
15 5. Empirical findings
This chapter presents the empirical findings of this research, distributed over three sections. Section 5.1 performs a univariate analysis on the stock price effect of football game outcomes and discusses the hypotheses that are reported in the literature review. Section 5.2 executes a multivariate analysis on the main findings of the univariate analysis, i.e. football game variables are combined to test whether these variables can explain the findings. Finally, Section 5.3 performs several robustness checks on the used variables and data.
5.1 Univariate analysis
This section is structured in the same sequence of the derived hypotheses in the literature review. This study first performs a univariate analysis to investigate the effect of football game outcomes on the stock market, by measuring the club’s stock returns on the first trading day after the game. The market model estimates expected return using the market index of the related country as a benchmark. Appendix D reports the average stock return and standard deviation of the clubs included in the sample over the seasons 2013-14 till 2016-2017. Moreover, Appendix D shows the alpha and beta parameters and their significance level that are estimated by the market model.
Hypothesis I: Positive football game outcomes result in positive abnormal stock returns of the club, while negative outcomes generate negative abnormal returns.
16 of draws and losses, and the higher impact of losses than wins are consistent with the findings in the literature (among others Dimic et al.,2018; Palomino, Renneboog, and Zhang, 2009; Scholtens and Peenstra, 2009). Therefore, this study’s empirical findings support the first hypothesis.
Hypothesis II: The stock price reaction to losses is larger than for wins.
The second hypothesis proposes that stock prices react more strongly from losses than wins. Table 5 illustrates a significant abnormal return of -149 bps after a loss, while a win generates an average abnormal return 61 bps. These results show that losses generate a stock price reaction that is more than twice as large as wins. This finding confirms the study’s outcomes of Renneboog and Van Brabant, (2000), Palomino, Renneboog, and Zhang (2009), and Benkraiem, Louhichi, and Marques (2009) and validates the second hypothesis.
Hypothesis III: Unexpected game outcomes cause a stronger stock price response than expected outcomes.
The third hypothesis proposes the presence of a surprise effect: the stock price reaction increases with the level of surprise, i.e. unexpected outcomes have a larger effect on stock prices than expected outcomes, which is based on the efficient market hypothesis. As addressed in Chapter 2 and 3, predicted game outcomes are based on betting odds and are divided in four groups: strongly expected to win (SEW), weakly expected to win (WEW), weakly expected to lose (WEL), and strongly expected to lose. Findings show mixed results of the surprise effect with respect to wins. Table 5 shows an abnormal return of 61 bps of a club’s stock price after a win. The abnormal return of a strongly expected win is lower (29 bps), while a weakly expected win is significantly higher (164 bps). However, a weakly unexpected win (WEL) causes abnormal returns of on average 99 bps. When focusing on losses, weakly expected losses generate a significantly higher stock price response than unexpected losses. Weakly expected losses cause significant abnormal returns of 210 bps, compared to a stock price response of -156 and -146 bps after weakly and strongly unexpected losses. It is ambiguous whether or not these disparities provide evidence to support or reject this hypothesis since only 36% of the weakly expected to lose games are indeed lost and strongly expected losses do not give a significant result. Strongly expected to lose games do also not give significant abnormal returns after a win or draw. These insignificant results are possibly due to a forced selection bias. The sample size for strongly expected to lose games is very small since listed football clubs are on average performing very well in the national competition.
17 losses decrease in absolute terms with the surprise level, excluding the insignificant strongly expected to lose games. These opposing results could be explained by other football game variables or could indicate the presence of a sentiment. Section 5.2 performs a multivariate analysis and tries to explain these results by involving other variables, such as goal difference and venue of the game. To summarize, these findings neither supports nor rejects the third hypothesis. However, the current study’s findings to some extent support the studies of Dimic et al. (2018) and Scholtens and Peenstra (2009), as the findings indicate that the stock market behaves irrational and investors' sentiment drives stock price reactions.
Table 5. Club’s average stock returns on the first t trading days after a football game.
Wins Draws Losses
Day 1 Day 2 Day 3 Day 1 Day 2 Day 3 Day 1 Day 2 Day 3 p-value p-value p-value p-value p-value p-value p-value p-value p-value
Panel A: Average abnormal return in % on day t
Total 0.61*** -0.02 0.16 -0.74*** -0.14 0.16 -1.49*** -0.13 -0.00 0.00 0.84 0.41 0.00 0.41 0.32 0.00 0.47 1.00 SEW 0.29** -0.16 0.01 -1.41*** -0.12 0.12 -1.46*** 0.05 -0.09 0.01 0.19 0.91 0.00 0.60 0.51 0.00 0.86 0.75 WEW 1.64*** 0.21 -0.03 -0.67** 0.10 0.21 -1.56*** -0.37 -0.13 0.00 0.40 0.90 0.02 0.78 0.50 0.00 0.19 0.72 WEL 0.99** 0.57 2.48 0.34 -0.31 0.37 -1.77*** -0.22 0.11 0.04 0.11 0.32 0.42 0.41 0.44 0.00 0.55 0.79 SEL 0.44 0.96 -1.37 1.14 -0.89 -0.57 -0.83 0.20 0.33 0.49 0.59 0.12 0.28 0.37 0.65 0.22 0.78 0.59
Panel B: Cumulative average abnormal return in % from day 0 to day t
Total 0.59*** 0.75*** -0.74*** -0.88*** -0.7*** -1.49*** -1.61*** -1.61*** 0.00 0.00 0.00 0.00 0.01 0.00 0.00 0.00 SEW 0.13 0.14 -1.41*** -1.52*** -1.40*** -1.46*** -1.41*** -1.50*** 0.41 0.48 0.00 0.00 0.00 0.00 0.00 0.00 WEW 1.85*** 1.82*** -0.67** -0.58 -0.37 -1.56*** -1.92*** -2.06*** 0.00 0.00 0.02 0.23 0.54 0.00 0.00 0.00 WEL 1.57*** 4.05 0.34 0.03 0.40 -1.77*** -1.99*** -1.88*** 0.00 0.12 0.42 0.96 0.56 0.00 0.00 0.00 SEL 1.40 0.03 1.14 0.25 -0.32 -0.83 -0.63 -0.30 0.49 0.99 0.28 0.82 0.86 0.22 0.43 0.69
Source: Datastream, Oddsportal.
Table 5 presents a football club’s (cumulative) average abnormal returns on the first trading day after a game, described as Day 1, categorized in the outcome of the game (win, draw, loss). Day 2 and 3 illustrate the two subsequent days. SEW, WEW, WEL, and SEL describe the expected outcomes based on bookmakers' betting odds (see Section 3.2), and define strongly expected to win, weakly expected to win, weakly expected to lose and strongly expected to lose, respectively.
18
Hypothesis IV: Negative football game outcomes are slower absorbed in the stock market than positive outcomes.
Hypothesis IV suggests that the market responds quicker to positive outcomes than to negative outcomes. Table 5 shows only significant abnormal returns on day 1, which indicates that the price response of positive and negative outcomes mainly appears on the first trading day after the game. The first trading day after losses experiences an average abnormal return of 149 bps, while the cumulative abnormal return on the third day after is 161 bps, which indicates that 92% of the market response occurs on the first day after a game. A win generates an abnormal return 61 bps on the first trading day, which is 81% of the cumulative abnormal return on the third day. This finding indicates that the market quickly absorbs the release of a football game outcome, regardless a positive or negative outcome. This result contradicts the study of Dimic et al. (2018), which argues that the second and third day after a game generates 60% of the total price response. These conflicting results can be explained by the smaller club sample size (13 clubs) that Dimic et al. use or by the different and larger sample period (2001-02 to 2012-13). Although the abnormal returns on the first three days after losses and wins are not significant, the cumulative abnormal returns are significant, which indicates that it takes at least three days for the stock price to recover from the loss effect or deteriorate from the win effect. An alternative explanation for these significant cumulative abnormal returns on day three is the impact of future economic effects on the long term that is already reflected in the stock prices. To summarize, this study’s results do not provide evidence for a slower stock market response for losses than wins and is therefore not able to support the third hypothesis.
Hypothesis V: Emotional load in a game causes (higher) abnormal returns.
19 than 30% higher than for ‘small’ losses. This finding is in line with the study of Dimic et al. (2018) and supports the fifth hypothesis.
Losses from home games have a high emotional load since clubs are in general better performing at home than away, hence a home loss would also indicate a loss away against the same club. The sample in this study generates a higher stock price reaction for won away games than home games, as presented in Table 6. However, the significant effect of a home draw or loss is stronger than an away draw or loss. Therefore, the result regarding the venue of the game only partly supports the fifth hypothesis. Moreover, this result slightly corresponds with the findings of the study of Dimic et al. (2018), Benkraiem, Louhichi and Marques (2009) and Benkraiem, Le Roy and Louhichi (2011) that experience stronger abnormal price reactions for home games, particularly for home losses according to the latter two studies.
Table 6. Football club’s stock returns in percentages after games for different panels.
Win Draw Loss
AAR (1) CAAR(1,2) AAR (1) CAAR(1,2) AAR (1) CAAR(1,2)
p-value p-value p-value p-value p-value p-value
Panel A: Goal difference
1 0.14 0.13 -1.65*** -1.90*** 0.355 0.518 0.000 0.000 2 0.90*** 1.01*** -0.95*** -0.98** 0.000 0.001 0.007 0.046 3 0.87*** 0.58* -2.16** -2.01** 0.002 0.094 0.014 0.020 > 3 1.00*** 0.82*** -1.90*** -1.63** 0.000 0.003 0.003 0.012 Panel B: Venue Home 0.40*** -0.01 -0.82*** -1.10*** -1.51*** -1.39*** 0.004 0.967 0.003 0.003 0.000 0.000 Away 0.90*** 1.39*** -0.68*** -0.71 -1.48*** -1.74*** 0.000 0.000 0.001 0.014 0.000 0.000
Panel C: Time in season
July-Oct 0.90*** 0.89*** -0.85** -0.62 -1.64*** -1.99*** 0.000 0.001 0.023 0.240 0.000 0.000 Nov-Feb 0.74*** 0.68*** -0.79*** -0.84** -1.33*** -1.23*** 0.000 0.003 0.005 0.012 0.000 0.002 March-June 0.31 0.38 -0.69*** -1.20*** -1.65*** -1.77*** 0.125 0.173 0.004 0.001 0.000 0.000
Panel D: Final standing
20
Win Draw Loss
AAR (1) CAAR(1,2) AAR (1) CAAR(1,2
) AAR (1) CAAR(1,2)
p-value p-value p-value p-value p-value p-value
Panel E: Club size
Small 0.78** 0.86** -0.04 -0.76 -1.80*** -1.83*** 0.020 0.043 0.921 0.203 0.000 0.000 Medium 0.62*** 0.59** -0.84*** -0.66** -1.19*** -1.33*** 0.000 0.017 0.000 0.036 0.000 0.000 Large 0.52*** 0.46*** -1.21*** -1.16*** -1.37*** -1.62*** 0.000 0.006 0.000 0.000 0.000 0.000 Source: Datastream.
This table shows a summary of a club’s average abnormal returns on the first trading day after a football game. Table 15-19 in Appendix E provide a broad overview of the abnormal returns. The whole sample of games is differently allocated for four panels. Panel A, B, C, and D refer to these different allocations. Panel A distributes the whole sample in groups that are based on the goal difference in the football game. Panel B splits home and away games. Panel C differentiates between the time in the season. The end of the season includes May and June, and the other months belong to the rest of the season. Panel D distributes the sample in two groups that are based on size. The small-sized group in consists of football clubs with a market value lower than €10 million, medium-sized clubs have a market value between €10 and €100 million, and the large-sized clubs have a market value higher than €50 million. AAR = Average abnormal return, p-value = probability value based on the Student’s t-statistic.
* Significant at the 10% level ** Significant at the 5% level *** Significant at the 1% level
89 and -188 compared to 53 and -132
Hypothesis VI: Games of higher importance generate stronger stock reactions than games of less importance.
21 Another variable that can influence the importance of the game is the club’s performance within a season, which is in this study measured by the final standing. Table 18 in Appendix E analyzes different categorizations of final standing groups and shows that the top three, stand four to six, and higher than six yield similar results. Panel D in table 6 shows the abnormal returns for these three groups. Both wins and losses have a larger effect on clubs with a standing lower than six (+89 and -188 bps respectively) than top three clubs (+53 and -132 bps respectively). However, as shown in the data section, club size and performance level are positively related in the observed sample. This indicates that these different results for the different performing clubs could also be explained by club size, which will be discussed in the subsequent hypothesis. Therefore, based on these results, it cannot be concluded that the final standing influences the degree to which stock prices response to football game outcomes.
Hypothesis VII: Small football clubs experience a stronger stock price reaction from games than large football clubs.
To test the previously mentioned hypothesis, the football clubs are divided into three samples, based on the average market value from the period 2013-2017 (see Section 3, Table 1 for the market values per club). The small group consists of six clubs that have a market value lower than €10 million. The seven medium-sized clubs have an average market value between €10 and €100 million and the other seven clubs in the large group have a market value higher than €100 million. Panel E in Table 6 shows the effect of football game outcomes on a club’s stock price for different club sizes. The abnormal return on day 0, as well as the cumulative abnormal return on day 1, is absolutely higher for small clubs than for large clubs that win or lose a game. For draws, large clubs experience a stronger reaction on day 0. However, the CAAR on day 1 is again in absolute terms higher for small clubs. As described in the literature review, existing studies argue that smaller stocks more strongly respond to events than large stocks (Baker and Wurgler, 2006) and Edmans, García, and Norli (2007). Considering findings of the current and previous studies, the seventh hypothesis is supported with respect to wins and losses.
5.2 Multivariate analysis
22 As shown in the previous section, weakly expected won football games generate higher abnormal returns (164 bps) than strongly expected won games (29 bps). Table 7 summarizes the abnormal returns after (un)expected losses and wins that are allocated based on the club’s size, period within the season, final standing, venue, or goal difference. Table 20a-d from Appendix F provides a complete overview of these results on the days subsequent to the game. The high abnormal return of a weakly expected win is mostly explained by the smaller clubs, on average 260 bps, while large clubs under the same conditions generate abnormal returns of 87 bps. The high negative stock price response of -177 bps after weakly expected losses is caused by small football clubs, which generate abnormal returns of on average -246 bps. The stock price effect of weakly and strongly expected losses is stronger at/around the start of the season, compared to the end. This effect can be caused by a sentiment effect. Public’s expectations of losing are confirmed by the game outcome, which could indicate more losses during the rest of the season. Strongly unexpected losses generate in the begin of the period an average abnormal return of -20 bps, while at the end of the season -179 bps. These losses could indicate a lower final standing of the period, since the end of the season is approaching. The significantly positive response of 99 bps after a weakly unexpected win is largely caused by games in the beginning of the season that generate abnormal returns of 240 bps. Strongly unexpected losses have a significantly stronger effect on clubs that have a final standing of seven or higher than a club that ends in the top three, i.e. a negative abnormal return of -361 bps for top three clubs compared to -152 for clubs that end at a seventh or lower stand. The insignificant abnormal returns from strongly expected to lose games cannot be explained by any of the investigated variables. This can be again due to the small sample size for (expected) losses. To summarize, weakly expected wins and losses are mainly caused by small-sized clubs and is larger in the begin of the season than in the end. Moreover, the negative effect after strongly unexpected losses is substantially larger at the end of the season and for clubs that have a final standing of 7 or higher.
Table 7. Football club’s stock returns in after games for different expectations and panels.
SEW WEW WEL SEL
23
SEW WEW WEL SEL
AAR Win AAR Loss AAR Win AAR Loss AAR Win AAR Loss AAR Win AAR Loss p-value p-value p-value p-value p-value p-value p-value p-value
Panel B: Period in season Begin 0.47** -0.20 1.55*** -1.79*** 2.45** -2.16*** 0.16 -1.86 0.03 0.74 0.00 0.00 0.01 0.00 0.47 0.13 End -0.07 -1.79*** 1.61*** -1.79*** -0.61 -1.52* 1.06 -1.30 0.74 0.00 0.01 0.00 0.35 0.05 0.32 0.28
Panel C: Final standing Top 3 0.23* -1.52*** 1.75*** -1.18** 1.37* -1.40*** 0.54 -0.07 0.05 0.00 0.00 0.01 0.09 0.01 0.68 0.94 7 + 1.07 -3.61*** 0.92 -1.11** 0.80 -2.32*** 0.50 -1.45 0.12 0.00 0.13 0.02 0.32 0.00 0.56 0.14 Panel D: Venue Home 0.14 -1.05** 1.71*** -1.80*** 1.67** -2.72** -0.35 1.71 0.28 0.03 0.00 0.00 0.02 0.01 0.52 0.56 Away 0.57*** -1.85*** 1.59*** -1.41*** 0.76 -1.33*** 0.66 -1.26** 0.01 0.00 0.00 0.00 0.22 0.00 0.42 0.05
Panel D: Goal difference
1 -0.05 -1.51*** 0.64** -1.94*** 0.01 -1.90*** 0.40 -0.53 0.79 0.00 0.04 0.00 0.99 0.00 0.54 0.66 2 0.35* -1.48*** 2.97*** -0.79 1.51*** -0.58 -2.69 -1.10 0.05 0.00 0.00 0.25 0.00 0.41 0.10 0.36 > 3 0.59*** -0.91 2.62*** -1.66 4.59 -3.24** 2.64 -0.95 0.01 0.47 0.00 0.16 0.10 0.02 0.13 0.41
Source: Datastream, Oddsportal.
SEW, WEW, WEL, and SEL describe the expected outcomes based on bookmakers' betting odds (see Section 3.2), and define strongly expected to win, weakly expected to win, weakly expected to lose and strongly expected to lose, respectively. The club’s size is based on its market value, where small clubs have a market value lower than €10 million and large ones higher than €100 million. The begin of a season includes the games in July until October, and the end includes March until June. AAR = Average abnormal return in % on the first trading day after a game, which could be a loss or win. P-value = probability value based on the Student’s t-statistic.
* Significant at the 10% level ** Significant at the 5% level *** Significant at the 1% level
24 and wins away with a high goal difference as very important, which could predict future game outcomes or show the presence of a sentiment effect.
A high goal difference for wins has more impact when clubs have a lower ranking. An extreme win (three or more goal differences) generates an average abnormal return of 89 bps for top three clubs, compared to 135 and 217 bps for clubs that are lower ranked at the end of the season. A high goal difference for losses also has more impact when have a lower ranking. To summarize, a football game with a high goal difference has more stock price impact on clubs that have a low final standing. Extreme losses have a significant effect small clubs’ stock prices, that generate substantially large negative abnormal returns of -276 bps. This similar effect for small and low ranked clubs is as expected, since small clubs are on average lower ranked than larger clubs. Moreover, extreme losses cause large negative stock price decreases of -336 bps at the end of a season, compared to -236 bps in the beginning of the season, which could be explained by a possible relegation or lower standing.
Table 8. Football club’s stock returns after games categorized on goal differences.
-3 to -7 -2 -1 0 1 2 3 to 6
AAR (0) AAR (0) AAR (0) AAR (0) AAR (0) AAR (0) AAR (0) p-value p-value p-value p-value p-value p-value p-value All games -1.90*** -0.95*** -1.65*** -0.74*** 0.14 0.90*** 1.00*** 0.00 0.01 0.00 0.00 0.35 0.00 0.00 Panel A: Venue Home -4.21** -1.17** -1.26*** -0.82*** 0.03 0.67** 0.62*** 0.01 0.04 0.01 0.00 0.90 0.03 0.01 Away -1.06* -0.82* -1.91*** -0.68*** 0.27 1.16*** 1.76*** 0.09 0.07 0.00 0.00 0.23 0.00 0.00
Panel B: Final standing
Top 3 -0.91 -0.74 -1.66*** -1.05*** 0.07 0.82*** 0.89*** 0.22 0.18 0.00 0.00 0.69 0.00 0.00 Stand 4 - 6 -1.91** -0.47 -1.56** -0.21 0.61 1.47*** 1.35** 0.05 0.46 0.03 0.69 0.29 0.00 0.04 7 + stand -2.79** -1.56** -1.69*** 0.12 0.31 0.92* 2.17** 0.02 0.02 0.00 0.80 0.56 0.09 0.04 Panel C: Size Small -2.76** -1.00 -1.87*** -0.04 0.28 1.23 1.19* 0.02 0.18 0.00 0.92 0.52 0.10 0.07 Medium -0.68 -0.57 -1.66*** -0.84*** -0.03 0.91*** 1.10*** 0.38 0.12 0.00 0.00 0.92 0.00 0.00 Large -1.74 -1.37** -1.39*** -1.21*** 0.22 0.73*** 0.82*** 0.12 0.02 0.00 0.00 0.26 0.01 0.00
Panel D: Period during season
25 This table shows the effect of goal differences in a football game from the perspective of the home team. To clarify, the category -3 to -7 include lost games with a goal difference of a number equal to or between three and seven. The club’s size is based on its market value, where small clubs have a market value lower than €10 million and large ones higher than €100 million. The begin of a season includes the games in July until October, and the end includes March until June. AAR = Average abnormal return in % on the first trading day after a game, which could be a loss or win. P-value = probability value based on the Student’s t-statistic.
* Significant at the 10% level ** Significant at the 5% level *** Significant at the 1% level
Overall, losses and wins from clubs that have a standing lower than six, cause a higher abnormal return than top three clubs. For the bottom standing clubs, games at the beginning of the period can be considered as more important. This could be due to the fact that the final standing is still uncertain and these clubs have the chance to have a high standing at the end of the season. The high negative abnormal return from a top three club’s loss, -132 bps, is mostly from large clubs, since these clubs have on average an abnormal return of -160 bps after a loss. Moreover, losses from a large club that has a final ranking of 7 or lower generate on average a negative abnormal return of -238 bps, where small clubs under the same conditions generate abnormal returns of -184 bps. Overall, the large effect of losses on clubs that have a final standing of 7 or lower, is mainly caused by large-sized companies and at the end of the season.
Table 9. Game importance, AAR in %
Top 3 stand Stand 4 – 6 7 + stand
AAR Win AAR Loss AAR Win AAR Loss AAR Win AAR Loss p-value p-value p-value p-value p-value p-value Total 0.53*** -1.32*** 1.06*** -1.24*** 0.89** -1.88***
0.00 0.00 0.00 0.01 0.02 0.00
Panel A: Period during season
Begin 0.81*** -0.67 1.05 -1.99* 1.52** -2.64*** 0.00 0.12 0.19 0.09 0.03 0.00 Half 0.72*** -1.83*** 0.67* -0.85 1.05* -0.64 0.00 0.00 0.05 0.13 0.07 0.38 End 0.18 -1.71*** 1.46*** -0.93 0.16 -1.99*** 0.44 0.00 0.01 0.13 0.81 0.01
Panel B: Club size
26 This table shows the impact of several variables after football matches on a club’s stock performance, categorized for the final standing at the end of a season. A club’s size is based on its market value, where small clubs have a market value lower than €10 million and large ones higher than €100 million. The begin of a season includes the games in July until October, and the end includes March until June. AAR = Average abnormal return in % on the first trading day after a game, which could be a loss or win. P-value = probability P-value based on the Student’s t-statistic.
* Significant at the 10% level ** Significant at the 5% level *** Significant at the 1% level
5.3 Robustness of the findings
This section performs several robustness checks on the construction of several variables and econometric issues. First, the construction of the prediction variables are verified. Appendix C and D show whether the effect of changing the variables that construct prediction variables SEW, WEW, WEL, and SEL affect this study’s findings. Section 3.2 presents the originally used sets of variables for specification [a] and [b], which are respectively {0.25,0.35,0.45} and {-0.3,0,0.3}. The alternative sets of variables for specification [a] are {0.2,0.3,0.4} and {0.3,0.4,0.5} and for specification [b] {-0.2,0,0.2}, {-0.25,0,0.25}, and {-0.4,0,0.4}. Appendix C shows the number and distribution of the expected and real game outcome, according to the bookmaker’s betting odds and based on the alternative estimate variables. The tables show that the percentage of right predicted outcomes is similar results shown in Section 3.2 for each alternative. Appendix D shows the average abnormal returns derived from the expected wins, draws, and losses for each prediction alternative. These results show similar abnormal returns and corresponding significance level for the used prediction variables in the study and alternative variables.
The second variable that needs to be verified is the post-February effect, which measures the (increasing) importance level of games played at the end of the season and thus a stronger stock price effect. Appendix E presents the impact of a win, draw and loss for different (groups of) months. The appendix shows that the abnormal returns do not change substantially when using a post-January or post-March as an indication for the end of the season. The only difference is that post-January wins generate a significant abnormal return.
27 heteroscedasticity is the generalized least squares (GLS). Other solutions for heteroscedasticity are using the variables’ logs to rescale the data or using White’s (1980) standard errors.
The second assumption that is tested is that the covariance of the error terms (abnormal returns) over time is zero. A covariance unequal to zero indicates that the error terms are ‘serially correlated’ or ‘autocorrelated’. Appendix F presents the probabilities of the Breusch-Godfrey serial correlation test. These results show that the null hypothesis of no autocorrelation is rejected for all clubs and thus proves the presence of autocorrelation. Ignoring the presence of autocorrelation yields the same consequence as ignoring heteroscedasticity and a possible solution is also using a GLS estimation method. Palomino, Renneboog, and Zhang (2009) use another method to adjusted their stock returns for autocorrelation. Their study applies the method of Brown and Hartzell (2001), which verifies the autocorrelation by an AR(1) and MA(1) model.
Thirdly, the assumption of a normally distributed residuals validated, which can be tested by the Bera-Jarque test. Appendix F shows the test-statistics that indicate non-normality at all clubs, except for Brondby. Abnormal returns that are not normally distributed are often tested using the non-parametric Corrado (1989) test. This test ranks all abnormal returns in the estimation window3. Appendix G shows the average abnormal returns after a win, draw, and loss and the corresponding Student t-statistic and Corrado rank statistic. The two types of test statistics show similar significance levels.
28 6. Conclusion
This research investigates the effects of football game outcomes on stock prices of 20 listed European clubs. Abnormal returns on the first trading day(s) after a game measure these effects. The average abnormal return is calculated as the average difference between realized and expected returns. The stock market index of a club’s country serves as a benchmark for the expected return of that club. This investigation uses several control variables to determine the effects of a particular type of game outcome on the stock price, which are the following: the outcome of the game, expectations, goal difference, venue of the game, importance and size of the football club. Bookmakers' betting odds derive a game’s outcome expectations. These betting odds reveal how much money a bettor will earn when he/she made the correct prediction. Games are distributed over four different expectation groups, which are based on the betting odds. The groups are the following: strongly expected to win, weakly expected to win, weakly expected to lose and strongly expected to lose.
29 abnormal returns in the last two months of a season than in the rest of the season. However, wins at the end of the season do not generate significant abnormal returns. The last hypothesis number seven is supported by the findings of this study since wins and losses produce higher abnormal returns in smaller clubs than in larger clubs.
The multivariate analysis analyzes different variables with respect to the three main variables from the univariate analysis: predicted game outcome, emotional load, and level of importance. When focusing on predicted game outcome, weakly expected wins and losses are mainly caused by small-sized clubs and is larger in the begin of the season than in the end. Moreover, the negative effect after strongly unexpected losses is substantially larger at the end of the season and for clubs that have a final standing of 7 or higher. Lost games with a high goal difference affect the club’s stock price substantially, which supports the findings of Benkraiem, Le Roy, and Louhichi (2011). Besides, extreme losses have a significant larger effect on the stock prices at the end of the season, compared to the begin of the season.
According to these findings, the answer to the primary research question which variables from football games can explain clubs' stock price reactions, is small clubs, clubs with a final standing of 7 or higher, games at the end of the season, and games with a high goal difference. Contrasting to the expectations, findings show ambiguous results regarding the effect of expected and unexpected outcomes.
Overall, the main findings suggest that wins, draws and losses affect the stock price respectively positive, to a small degree negative, and substantially negative on the first trading day after a game. A higher surprise level in win results in higher abnormal returns. In addition, both a larger goal difference and a smaller club produce a stronger stock price reaction. This study contributes to the literature as it confirms the main findings of previous studies of e.g. Dimic et al. (2018), while using a different and larger sample. Moreover, this study investigates many possible game variables that can influence the response of a club's stock price. Last, the results of the current study show that both positive and negative game results are directly absorbed in the market, while studies with an older dataset show that negative outcomes are slower absorbed by the market than positive outcomes (see, e.g. Palomino, Renneboog, and Zhang, 2009). These contrasting results could indicate that there is a transition in the pace of absorbing information in stock prices. Therefore, a further study could assess the pace of absorbing information for other negative announcements.
30 Section 5.2 provides several solutions for these problems. Future research could also include international games in the sample.
31 References
Ashton, J.K., Gerrard, B. and Hudson, R., 2003. Economic impact of national sporting success: evidence from the London stock exchange. Applied Economics Letters, 10(12), pp.783-785.
Baker, M. and Wurgler, J., 2006. Investor sentiment and the cross‐section of stock returns. The Journal of Finance, 61(4), pp.1645-1680.
Barajas, A., Fernández-Jardón, C.M. and Crolley, L., 2005. Does sports performance influence revenues and economic results in Spanish football? Unpublished working paper. University of Vigo, Vigo, Spain.
Bell, A.R., Brooks, C., Matthews, D. and Sutcliffe, C., 2012. Over the moon or sick as a parrot? The effects of football results on a club's share price. Applied Economics, 44(26), pp.3435-3452.
Benkraiem, R., Le Roy, F. and Louhichi, W., 2011. Sporting performances and the volatility of listed football clubs. International Journal of Sport Finance, 6(4), p.283.
Benkraiem, R., Louhichi, W. and Marques, P., 2009. Market reaction to sporting results: The case of European listed football clubs. Management Decision, 47(1), pp.100-109.
Bernile, G. and Lyandres, E., 2011. Understanding investor sentiment: The case of football. Financial Management, 40(2), pp.357-380.
Berument, H., Ceylan, N.B. and Gozpinar, E., 2006. Performance of football on the stock market: Evidence from Turkey. The Social Science Journal, 43(4), pp.695-699.
Berument, H. and Yucel, E.M., 2005. Long live Fenerbahçe: The production boosting effects of football. Journal of Economic Psychology, 26(6), pp.842-861.
Brooks, C., 2002. Introductory Econometrics for Finance. Cambridge University Press, Cambridge.
32 Corrado, C.J., 1989. A nonparametric test for abnormal security-price performance in event studies. Journal of financial economics, 23(2), pp.385-395.
Dimic, N., Neudl, M., Orlov, V. and Äijö, J., 2018. Investor sentiment, football games and stock returns. Research in International Business and Finance, 43, pp.90-98.
Duque, J. and Ferreira, N.A., 2005. Explaining share price performance of football clubs listed on the Euronext Lisbon. Unpublished working paper. Technical University of Lisbon, Lisbon.
Edmans, A., Garcia, D. and Norli, Ø., 2007. Sports sentiment and stock returns. The Journal of Finance, 62(4), pp.1967-1998.
Falconieri, S., Palomino, F. and Sákovics, J., 2004. Collective versus individual sale of television rights in league sports. Journal of the European Economic Association, 2(5), pp.833-862.
Forrest, D., Goddard, J. and Simmons, R., 2005. Odds-setters as forecasters: The case of English football. International journal of forecasting, 21(3), pp.551-564.
Forrest, D. and Simmons, R., 2000. Forecasting sport: the behaviour and performance of football tipsters. International Journal of Forecasting, 16(3), pp.317-331.
Hvattum, L.M. and Arntzen, H., 2010. Using ELO ratings for match result prediction in association football. International Journal of forecasting, 26(3), pp.460-470.
Kaplanski, G. and Levy, H., 2010. Exploitable predictable irrationality: The FIFA World Cup effect on the US stock market. Journal of Financial and Quantitative Analysis, 45(2), pp.535-553.
MacKinlay, A.C., 1997. Event studies in economics and finance. Journal of economic literature, 35(1), pp.13-39.
Morrow, S., 1999. The New Business of Football: Accountability and Finance in Football. Palgrave Macmillan, Basingstoke.
33 Palomino, F. and Sákovics, J., 2004. Inter-league competition for talent vs. competitive balance. International Journal of Industrial Organization, 22(6), pp.783-797.
Renneboog, L. and Vanbrabant, P., 2000. Share price reactions to sporty performances of soccer clubs listed on the London Stock Exchange and the AIM. Unpublished working paper. Tilburg University, Tilburg.
Saraç, M. and Zeren, F., 2013. The Effect of Football Performance on Stock Return: Empirical Evidence From" The Big Three Clubs" of Turkish Football League. Journal of Applied Finance and Banking, 3(5), p.299.
Scholtens, B. and Peenstra, W., 2009. Scoring on the stock exchange? The effect of football matches on stock market returns: an event study. Applied Economics, 41(25), pp.3231-3237.
Spann, M. and Skiera, B., 2009. Sports forecasting: a comparison of the forecast accuracy of prediction markets, betting odds and tipsters. Journal of Forecasting, 28(1), pp.55-72.
Stadtmann, G., 2006. Frequent news and pure signals: The case of a publicly traded football club. Scottish Journal of Political Economy, 53(4), pp.485-504.
