Football Sponsorship and Stock Market Returns

(1)

Football Sponsorship and Stock Market Returns

Studentnr.: S2814056 Student: Arjan Postma Study Program: MSc Finance Thesis supervisor: Dr. G.T.J. Zwart

Field Key Words: Football sponsorship; stock market efficiency; event study; investor sentiment

Abstract

This thesis examines whether football game results in the Champions league can affect the stock prices of jersey sponsors. It is established that this is indeed the case. Notably, a loss that was unexpected ex-ante, based on objective betting odds, leads to an abnormal return of -77 basis points on the next trading day. When an unexpected loss occurs in an

elimination round match, the abnormal returns further decrease to -172 basis points. The loss effect found is robust to the omission of outliers and is believed to have an economic cause rather than a behavioural.

I. Introduction

This study examines the effect of football game results in the Champions League on the stock prices of the corresponding jersey sponsors. The study was conducted by using an extensive dataset, covering ten seasons of Champions League football over the period 2007-2017 and includes the main jersey sponsors of all teams participating in the tournament. It was subsequently investigated whether wins and losses are associated with positive (negative) stock returns on the first trading day after a match was played.

The main null hypothesis is that football results do not impact the stock price of a sponsoring company. Under this null hypothesis, it is assumed that markets are semi-efficient and that rational investors perceive the economic benefits of a certain football team winning a Champions League match to be too small to influence the stock price of the corresponding jersey sponsor.

The motivation for writing a thesis on this subject was induced by a body of recent literature that investigated the relation between football results and stock returns. For example, Ashton et al. (2003) and Edmans et al. (2007) sought to link football results in international tournaments to an overall countrywide stock market reaction. On the subject of football results and stock market implications for sponsoring firms, Hanke & Kirchler (2013)

investigated the consequences of football results in international tournaments on the jersey sponsors.

This research seeks to elaborate on the aforementioned studies by exploring whether a significant relationship can be established between football results and the financial

(2)

with a football club, by identifying important potential opportunities and dangers of such an agreement.

The findings of this paper indicate a significant decline of 77 basis points in the stock prices of jersey sponsors, one trading day after an unexpected loss by the football club that is sponsored. Notably, when the unexpected loss occurred in a knockout match, the decline amplifies to 172 basis points. No significant effects are found for losses that were expected and for victories altogether.

Primarily based on the findings that Champions League match results only affect jersey sponsors when the match outcomes are unexpected and on the results being of a larger magnitude in knockout matches, it is argued that the found effects can be contributed to economic reasoning. That is to say, investors perceive the foregone exposure benefits from not progressing to a further stage in the tournament to be economically significant to the extent that it is reflected in the jersey sponsors’ stock prices.

The remainder of this thesis is structured as follows. The next section reviews the previous literature on the subject and outlines the hypotheses of this research. Section III presents the data that was used to conduct this study. Section IV describes the methods used to scrutinize the data. Section V reports the findings of this study and discusses them. Section VI concludes.

II. Literature Review & Hypotheses A. Previous literature

This research seeks to contribute to an existing body of literature on the subject of sports performances as a variable that could potentially impact asset prices. In the first decennium of this century, three relevant studies on the subject were published. Notably, Ashton et al. (2003) examined the effect that the results of the English national football team have on the FTSE 100 (The Financial Times stock exchange) one day after a match has been played. The authors conclude that there is a significant relationship between national football matches and national stock market returns. The day after a win by the English national team, the FTSE 100 displayed significant positive abnormal returns, while the returns were significantly negative in case of a defeat. The authors found that elimination matches in tournaments had a far greater influence than less important matches, such as friendlies. The authors

(3)

Edmans et al. (2007) reported similar results for the other sports they investigated (cricket, rugby and basketball) but these results are considerably smaller in magnitude.

Renneboog & Vanbrabant (2000) examined the effect of match outcomes on a football clubs’ own stock performance. Their research was focused on those clubs in the English Premier League which had their shares publicly listed at the time of their research. They found that wins resulted in positive abnormal returns of almost 1%, while draws as well as defeats led to negative abnormal returns of 0.6% and 1.4%, respectively. The results were stronger for those clubs listed at the London Stock Exchange than for the clubs listed at the Alternative Investment Market. Game importance was again found to be a crucial factor, since vital relegation games and important European matches yielded greater abnormal returns. Notably, positive and negative abnormal returns of around 3% were found. The authors attributed this increase to the fact that crucial games generate larger economic benefits for the football clubs in terms of broadcasting rights and sponsor income.

The studies above already indicate that football results can be significantly visible in financial markets. This effect can be direct when a club itself is publicly listed, as was the case in the work of Renneboog & Vanbrabant (2000). In that case the abnormal returns are a result of larger economic benefits for the winning team and vice versa when the team loses. On the other hand, Edmans et al. (2007) contributed the abnormal returns that they found after a win or a loss to a behavioural factor rather than an economic one. They argued that the results of the national football teams indirectly affected the performance of the particular country’s stock market, through the change in the mood of its investors following a win or a loss. Edmans et al. (2007) outlined three main arguments in favour of the behavioural explanation. Firstly, they found large effects for losses that were expected ex-ante, which cannot be reconciled with a semi-efficient market. Secondly, they found a larger effect for elimination games. The authors also attributed this to sentiment trading, as they believed that emotions amplify as matches become more crucial. Finally, Edmans et al. argued that the stronger effect observed for small stocks signifies sentiment trading. The link between small cap stocks and sentiment trading was already documented prior to their work by Baker & Wurgler (2006).

The studies presented above identified a few factors that altered the significance of the results. First and foremost, all three studies reported that the results depended heavily on the importance of the match that was played. Additionally, Renneboog and Vanbrabant (2000) found that their results depended on the stock exchange on which the football clubs in their sample were listed.

More recently, two further studies have been published that linked football results to corporate sponsorship. Hanke & Kirchler (2013) examined the effect of football matches on the stock market, but with the specific focus on the financial performance of the jersey sponsoring companies rather than the country’s stock market as a whole. The sample in this research consisted of nations participating in the World Cup and the European

(4)

This is in line with the “allegiance bias”, which was already documented by Edmans et al. (2007). They further confirmed the relevance of game importance and reported that the defeat effect, appeared more pronounced for important games (knockout games) than for less important matches (group stage games). Hanke & Kirchler (2013) further reported findings that were not yet discussed in any of the aforementioned articles. In addition to the loss effect, the authors reported a mere exposure effect, when two teams with the same sponsor played against each other on a given day. This effect was again more pronounced after more important games. Interestingly, the mere exposure effect found contradicts the efficient market hypothesis, pioneered by Burton & Fama (1970). The efficient market hypothesis articulates that in a semi-efficient market, all public information is reflected in stock prices. Since it is always known that two sponsors will be exposed simultaneously before the match is played (match fixtures are scheduled prior to the start of any match) the mere exposure effect is expected to be reflected in asset prices already. Therefore, the mere exposure effect seems to be based on sentiment rather than economic explanations. A final conclusion drawn in this paper concerned the existence of an inverse relation between the defeat effect and the ex-ante probability of a defeat. In other words, the loss effect is

stronger when the event (a loss) was unexpected, which is in line with a semi-strong efficient market. The expectations are based on objective betting odds. Hanke & Kirchler (2013) found that the amplified loss effect can be contributed to revised expectations by agents upon a loss that was not predicted by the betting odds. This conclusion leads to the belief that the loss-effect is the outcome of an economic mechanism. Notably, rational agents perceive the foregone economic benefits, due to elimination, to be so considerable that it influences the stock prices of the sponsors.

Similar findings were reported by Eisdorfer & Kohl (2015) in their analysis of the effect of American Football matches on stock returns of sponsoring companies in the NFL (National Football League), the American Football league in the United States. In their sample sponsoring concerned advertising inside the stadiums, rather than jersey sponsoring. Therefore, only home games were considered relevant in measuring the effect of NFL game results on the sponsors of the NFL teams. Eisdorfer & Kohl (2015) reported a difference of 50 basis points in the abnormal returns between wins and losses. The difference increases to 82 basis points when it concerns a relatively more important post-season match which

subsequently yielded an unexpected outcome. As no similar effect was documented for away games, Eisdorfer & Kohl (2015) suggest their results to be partially driven by sentiment trading.

B. Research hypotheses

This research seeks to add to the above mentioned literature by researching whether football results ignite effects on the stock returns of the companies that appear on the football teams’ kits as the official main jersey sponsor. The research focuses specifically on the Champions League, which is the largest European club tournament in football. What is more, with an estimated 350 million viewers all over the world, the Champions League final is the single biggest sport event in the world.1 Notably, only the main jersey sponsor, which appears on the front of the official kit of every football club, is considered in this research, as this is the one that receives the most prominent exposure.

(5)

The main hypothesis of this thesis is phrased as follows:

H1: Football results do not significantly impact the stock price of a sponsoring company It is important to clearly outline at this point that a potential effect can have an economic as well as a behavioural explanation as also suggested in various papers in the literature review. Under the economic mechanism, it is assumed that rational investors perceive a positive result in the Champions League as economically beneficial for the jersey sponsor. The reason for this perception is that a win in the Champions League increases the probability of

progressing to a further stage of the tournament which in turn leads to more exposure for the jersey sponsors. This increased exposure is twofold, i.e. added matches and increasing numbers of spectators. In the Champions League all participants start in a group stage, consisting of six fixtures. Upon progressing to the knockout phases, at least another two matches will be played, which would not have been the case had the team not survived the group stage. In other words, progressing to the knockout stage of the Champions League adds another minimum of 180 minutes of exposure for the jersey sponsor. The same will happen once again, every time the team survives another round of the knockout stage, with exception of the final which adds only 90 minutes of exposure. However, it is not solely the added exposure time that increases the exposure for the sponsors. Also, knockout matches in general generate larger audiences of viewership than group games, which further enlarges the exposure effect. Moreover, viewership generally increases further for every round in the knockout stage, with the final generating the largest audience. If investors believe that this increased exposure will positively impact sales and thus profits for the sponsoring company, the effect of a win should be visible in the stock price of the respective company. Naturally, a loss would instigate a reversed effect.

A second mechanism through which stock prices of jersey sponsors could be influenced is a behavioural mechanism, in which football results serve as a mood variable. The intuition here is that a positive result by a football team will be associated by investors with the jersey sponsor’s prospects. In other words, positive results by a football club ignite a similar

positive feeling towards the company displayed on that particular club’s kit, labelled by Ashton et al. (2003) as the “feel-good” factor. This associating is not economically rational. However, irrational associations between two unrelated variables have been proven before to influence investors, as was the case in the study by Hirshleifer & Shumway (2003). Notably, it was established that sunny weather causes an increase in stock prices. Similarly, when observing a team winning and cheering with the sponsoring company prominently exposed, similar positive feelings are ignited towards this specific company. Through this mechanism investors do not influence the price of a jersey sponsors by means of rational trading based on future economic benefits, but rather through sentiment trading.

Aside from investigating the main hypothesis, this research aims at distinguishing the two mentioned mechanisms. Therefore the following hypotheses are also examined (H2-H6), which are constructed to investigate whether any effect is detectable in the first place and to distinguish the aforementioned mechanism.

(6)

Hypothesis 2 is derived from findings by Edmans et al. (2007), where abnormal negative returns were reported after losses but there were no visible positive abnormal returns after victories. The authors attribute these results to an “allegiance bias”, in which investors are psychologically invested in a particular outcome. Plainly explained, the investor (considered a football fan) subjectively believes that his or her favourite team should be able to beat every opponent that they come up against and considers every loss as unexpected. As such, they overestimate their team’s probability of winning and as such reduce the actual effect of a win, while amplifying the effect of a loss. Since none of those beliefs concern the economic outlook of the company that serves as the football club’s jersey sponsor, confirmation of the allegiance bias in this research is considered a good indicator for the behavioural

mechanism.

H3: There is no difference between the effect following important matches (knockout games) and the effect following relatively less important matches (group games)

Hypothesis 3 is based on the studies conducted by Edmans et al. (2007) and Hanke & Kirchler (2013). If there is a clear difference between the results for important and less important games, it can be argued that the effect on the jersey sponsors’ stock price is a result of the economic mechanism. Champions League groups are decided over six games and therefore one win or one loss during the group stage need not have any profound consequences. This is different however in the knockout stage of the tournament. Although all rounds (with exception of the finals) are decided over two matches, a loss greatly

increases the possibility of elimination. Elimination in turn leads to foregone economic benefits, in the form of 180 minutes of exposure for the jersey sponsor.

Conversely, when there is no significant difference in the effect between important and less important games, this is a good indication that investors affect jersey sponsors’ stock prices through a behavioural mechanism.

H4: There is no difference between the effect following matches with unexpected outcomes and the effect following matches with an expected outcome

A similar hypothesis to H4 was researched by Edmans et al. (2007) and Hanke & Kirchler (2013). The argument behind this hypothesis is that in a semi-efficient market, expectations based on public information are already calculated into asset prices. In other words, when assuming the market to be semi-efficient, the economic value of a victory, when it was expected ex-ante, should already be reflected in the asset prices. If the result is not in accordance with expectations, these prices need revision. Therefore, unexpected wins and losses should produce stronger results than wins and losses that were already expected, based on objective probabilities (betting odds). Ergo, should expected outcomes also yield significant abnormal returns, than that would strongly suggest sentiment trading by

investors. This would favour the theory that the jersey sponsors’ stock prices were affected by the investors’ mood following football rather than because of economic implications.

(7)

stronger potential for local sentiment. A stronger effect for these stocks would thus indicate investor sentiment rather than trading for economic reasons. Since the small cap and large cap stock comparison is not feasible in the particular dataset of this research (since nearly all of the companies in the sample are large cap stocks), the locality of the jersey sponsor will be subject to scrutiny. Notably, a jersey sponsor is considered local when it is headquartered in the same country as the sponsored club originates from. The assumption here is that when a club is sponsored by a local company, its investors will have a stronger affinity with the football club. Moreover, they are more likely to invest in the jersey sponsor in the first place. The reasoning behind the latter stems from the “Home bias” phenomenon,

introduced by French & Poterba (1991). Home bias refers to the fact that investors are prone to invest close to home and therefore hold considerable fractions of local stocks in their portfolios, i.e. higher than what should actually be hold in an optimally diversified portfolio. Home bias may lead to a larger effect in a subsample of local companies, which would support the behavioural mechanism as the source of the abnormal returns.

H6: There are no significant abnormal returns when only finals are included

This sixth and final hypothesis has the potential of providing additional significant evidence in favour of the behavioural explanation. The reasoning behind this hypothesis is that if a team wins the final, there are no added exposure benefits for the jersey sponsor for the simple reason that there are no more additional matches after the finals. Hence, if a significant effect is visible after a win or loss in the finals, there would be no economical explanation for the significant abnormal returns. Ergo, support is found in favour of the explanation that the stock price of the jersey sponsor was affected by sentiment trading.

If this research would identify significant impacts on jersey sponsors, this has important implications for corporate sponsorship in general. Notably, this could create both dangers as well as opportunities for those companies that consider entering into a sponsorship

contract. Obviously, many factors come into play when deciding on a sports sponsorship agreement. However, it may not be the case that any brand exposure is positive exposure. Therefore, betting on the right horse (in this case a football club) could be vital and

companies should critically assess the chances of the team that they are considering to sponsor in the Champions League. In case a team fails to live up to the expectations it is possible that the sponsor will suffer, either because investors do not believe the sponsorship deal to be a positive net present value project or because of sentiment trading. Conversely, a team that exceeds expectations could prove an investment that is substantially more

(8)

III. Data

For the purposes of this research, the final outcomes of all Champions League fixtures between 2007 and 2017 were collected from the official UEFA Champions League website (https://www.uefa.com/uefachampionsleague), which comprises to a total of ten full

campaigns. Only group games and knockout games were collected, qualifying rounds are not accounted for in this research. Although including the qualifying rounds could potentially enlarge the dataset considerably, the exposure that these games receive is considered too minimal to contribute to answering the research questions.

A Champions League group consists of four teams and each team will play against each opponent twice, so every participating team plays six matches in the group stage. If a team finishes within the top two of their group, they will automatically be qualified for the knockout stage of the tournament. Knockout games are decided over two legs; therefore draws are technically a possibility in both matches. For the sake of this research the possibility of a draw in the decisive second leg of a knockout match has been eliminated. Despite the fact that the score can be tied after the second game, one team will progress to the next round (which is treated as a win) while the other team is eliminated from the tournament (which is treated as a loss). Similarly, it is possible for a team to lose in the second leg of a Champions League knockout match, but still progress because of the result in the first leg. When a situation as such arises, the result is still coded as a win because the team does progress to the next stage of the tournament. Conversely, a win in the second leg that was insufficient for a team to progress to the next round will be coded as a loss. The final is always decided in one single game, therefore wins and losses are always coded as such and naturally there is no possibility for a draw.

In order to measure the strengths of both teams and objectively quantify the ex-ante

probabilities of match outcomes, the website oddsportal.com (http://www.oddsportal.com) was consulted. On this website, all historic Champions League results are collected. Football odds are expressed in numbers (usually varying from 1 to 25), that represent the multiple of which you earn back your initial investment. Like investing in general, more risky

investments yield greater returns and vice versa, therefore the result that is favoured will have the lowest odd assigned by the bookmakers. Betting odds can easily be transformed into probabilities. For example, consider a well performing football team that is deemed advantageous and receives a betting odd of 1.25 from the bookmakers. In this research, a result is considered unexpected if the outcome of a match is not coherent to the result that was indicated by the lowest betting odd. One point that needs to be addressed is the case of an elimination match (second leg match of a knockout round), where a win (loss) can be recorded as a loss (win), when the team did not (did) proceed to the next round. The betting odds indicate the likelihood of a particular match outcome, therefore a win is always a win, a loss is always a loss and a draw is always a draw. However, given that bookmakers spend considerable time to inspect many factors that could influence a football match result (such as a team’s form and injuries to important players), it is also assumed that bookmakers take the results from the first leg into account and incorporate these results into the betting odds for the second leg of a knockout round.

(9)

For the majority of the European football clubs, sponsor contracts are short lived and

sponsors frequently replace each other. By means of the official websites of the participating football clubs and the website www.oldfootballshirts.com the corresponding jersey sponsors could be recorded for each participating club in the ten seasons. A prerequisite for the jersey sponsor to be included in the sample was that the company had to be publicly listed at the time they were sponsoring a particular Champions League contender.

In total, 54 different companies are accounted for in this research. A complete list of all teams and their corresponding jersey sponsors can be found in a table in appendix A. Finally, financial data on the companies was gathered from Yahoo finance

(https://finance.yahoo.com). All stock prices for all the companies included in the sample were generated from this database. In analysing the stock returns of the shirt sponsors, the Eurostoxx 50 index is used as a benchmark for European enterprises, while the S&P 500 index is used for non-European companies. The latter is a slightly controversial choice, since not all non-European firms in the sample are US based. However, due to limited data

availability for some of the other indices, this decision was made. A total of 1236 events are recorded over 300 match days in the sample, 508 games are wins, 478 are losses and 250 games are draws.

IV. Research methods

In order to analyse the data and investigate the hypotheses, a two-step study is required. The first step of the analysis consists of performing an event study, in order to obtain the abnormal returns for each of the 1236 events. The event study methodology used in this research follows the market model procedure, as outlined in MacKinlay’s paper (1997) on conducting event studies in economics and finance. The outcomes of the event study analysis provide a tentative indication of the effects that football results have on jersey sponsors. Yet, no inferences will be made from the event study outcomes as such.

The abnormal returns primarily form the input for the next step in the analysis, consisting of a regression analysis, in which the abnormal returns generated from the event study analysis are regressed as the dependent variable. The independent variables in the regression

equations concern the football results. Additionally, more specific independent variables attempt to distinguish the economic and behavioural mechanisms.

Step 1: Event study

Equation 1 presents the calculation used to obtain the daily stock returns for all of the collected stock prices.

𝑅𝑖,t =(𝑃1−𝑃0)_𝑃0 (1)

Where 𝑅𝑖,t is the return on stock i at time t on any given day, P1 is the ending stock price and

P0 is the initial stock price. Important to mention here is that P0 and P1 are adjusted close prices, which are adjusted for dividends and stock splits.

(10)

In order to test the normal performance of the stock, the market model prescribes the following linear equation:

E(𝑅_i,t) = α_𝑖 + β_𝑖R_m,t+ ε_i,t (2)

Equation 2 yields the expected returns [E(𝑅𝑖,t )], based on a linear regression that captures

the fluctuations of the market index (Rm,t). α𝑖and β𝑖 are the parameters of the market

model. εi,t reflects the random error term.

The final step taken to calculate the abnormal returns is to subtract the expected return from the actual return on the stock as formula 3 depicts:

𝐴𝑅_𝑖,𝑡 = 𝑅_𝑖,t− E (𝑅_𝑖,t) (3)

Ergo, the abnormal return (𝐴𝑅𝑖,𝑡) is the difference between the actual return of the stock

and the expected return of the stock, based on the performance of the market index.

Abnormal returns can be captured and accumulated over any number of days, depending on the length of the event window. Based on previous literature, the abnormal returns are collected only the first trading day after a match is played. There are time zone differences between the locations of the companies' headquarters included in the sample. Therefore the stocks of some companies could be traded upon or shortly after a Champions League match was played. Despite these asymmetries, abnormal returns were consistently captured one trading day after a match was played, so the possible effects have sufficient time to become apparent in every time zone.

Besides the event window, an estimation window is required as a point of reference. The estimation window is a period during which no events happen and the normal returns of a given firm at a given time can be estimated. Due to the occurrence of many events during the Champions League season, roughly every two weeks, the estimation window was chosen to be outside the season. Notably, the estimation window is comprised of the months prior to the start of each Champions League season (July and August) and the month after each Champions League season had ended (June).

Step 2: Regression analyses

The regression models, applied to the abnormal returns, in this section are predominantly based on Edmans et al. (2007) and Hanke and Kirchler (2013). The dependent variable for each of the following regressions is the abnormal return measured for company i at time t in step 1.

Equation 4 displays the regression formula that is applied to capture the general win and loss effect.

𝐴𝑅𝑖,𝑡 = α + β1𝑊𝐼𝑁𝑖,𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡 + εi,t (4)

(11)

In order to capture the effect of the match importance equation 4 is slightly adjusted. 𝐴𝑅𝑖,𝑡 = α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡 + εi,t (5)

𝐴𝑅_𝑖,𝑡 = α + β1𝑊𝐼𝑁_𝑖,𝑡∗ 𝐾𝑂_𝑡+ β2𝐿𝑂𝑆𝑆_𝑖,𝑡∗ 𝐾𝑂_𝑡 + ε_i,t (6)

Equations 5 and 6 both are an extended version of equation 4, with the addition of a binary dummy variable (GROUP or KO) which represents a group or a knockout game respectively. The value of the GROUP (KO) dummy will be 1 when the match played is a group (knockout) match and zero otherwise. By discerning these matches from each other, evidence could be obtained for the third hypothesis, which determines the cause of the abnormal returns. Thirdly, in order to determine whether the effect is more pronounced when results are unexpected, equation 7 and 8 need to be slightly altered in order to account for the pre-match odds in the base case regression and the regressions separating knockout pre-matches. 𝐴𝑅_𝑖,𝑡 = α + β1𝑊𝐼𝑁_𝑖,𝑡∗ 𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷_𝑖,𝑡+ β2𝐿𝑂𝑆𝑆_𝑖,𝑡∗ 𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷_𝑖,𝑡 + ε_i,t (7) 𝐴𝑅𝑖,𝑡 = α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑖,𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡 ∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑖,𝑡 + εi,t (8)

By comparing equations 7 and 8, it can be established whether the abnormal returns are more significant when a result was anticipated ex ante. The products of these regressions will determine whether or not hypothesis 4 can be rejected. When there is no evidence to reject this hypothesis, than that would be in favour of the behavioural explanation for the abnormal returns.

Equations 9 and 10 combine hypotheses 3 and 4 by comparing the abnormal returns from unexpected results in group matches and unexpected results in knockout matches.

𝐴𝑅𝑖,𝑡 α + β1𝑊𝐼𝑁𝑖,𝑡 ∗ 𝐺𝑅𝑂𝑈𝑃𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑖,𝑡 + β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡∗

𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷_𝑖,𝑡+ ε_i,t (9)

𝐴𝑅𝑖,𝑡 = α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐾𝑂𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑖.𝑡 + β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐾𝑂𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑖,𝑡+

ε_i,t (10)

When the results from Equation 9 and 10 yield more significant results than those from equations 7 and 8, than that would present strong evidence for the economic explanation of the abnormal returns. As that would indicate the relevance of match importance.

Equation 11 outlines the regression equation with a dummy variable which separates match results by teams with a local sponsor from the full sample:

𝐴𝑅𝑖,𝑡 = α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐿𝑂𝐶𝐴𝐿𝑖 + β2𝐿𝑂𝑆𝑆𝑖,𝑡 ∗ 𝐿𝑂𝐶𝐴𝐿𝑖 + εi,t (11)

Separating the local sponsors from the complete set of observations indicates whether there is evidence to reject hypothesis 5. If such evidence is found, than that would be in favour of the behavioural explanation of the abnormal returns.

Finally, equation 12 presents the regression model that involves a dummy with the intention of isolating the finals in the sample.

(12)

As specified in hypothesis 6, isolating the finals in the sample could yield evidence in favour of the behavioural explanation of the abnormal returns, in case the win and loss dummy variables turn out significant.

V. Results

A. Descriptive statistics

Table 1 summarizes the outcomes of the event study. The abnormal returns are the excess returns generated by the companies in the sample over a broad market index. As

(13)

Table 1.

Total number of Champions League games, knockout games and games with unexpected results and mean abnormal returns one trading day after a game.

This table reports the total amount of wins and losses in all Champions League matches from 2007 until 2017. Furthermore, the average aggregated abnormal returns collected from the event study analysis and their standard deviations are reported. All abnormal returns are recorded the first trading day after a match is played and benchmarked against a broad market index. The table also reports the amount of wins and losses and the abnormal returns for a subsample of group and knockout matches, a subsample of expected outcomes and matches with an unexpected outcome, a subsample of local sponsors and a subsample of finals. The mean abnormal returns presented in the table are not tested for significance.

All Wins Losses

All games N 1236 508 478 Mean 0.0029 0.0004 0.0049 SD 0.0811 0.0264 0.1249 Group games N 947 384 350 Mean 0.0040 0.0014 0.0072 SD 0.0916 0.0276 0.1457 Knockout games N 289 124 129 Mean -0.0004 -0.0021 -0.0019 SD 0.0246 0.0221 0.0226 Expected results N 758 398 344 Mean 0.0045 0.0010 0.0084 SD 0.1009 0.0271 0.1466 Unexpected N 478 110 135 results Mean 0.0004 -0.0005 -0.0051 SD 0.0296 0.0234 0.0235 Local sponsors N 606 245 246 Mean 0.0004 0.0006 -0.0004 SD 0.0182 0.0184 0.0180 Finals N 11 3 8 Mean 0.0032 0.0162 -0.0032 SD 0.0152 0.0177 0.0153

As can be seen in the table, the complete dataset comprises of 1236 events. 508 of those events are victories and result in an average abnormal return of 4 basis points. Over the whole sample, 478 losses surprisingly also result in an even higher positive average abnormal return of 49 basis points.

(14)

a win is -21 basis points. Similarly, a loss effect does seem to be present, since losses overall lead to a negative return of 19 basis points. So, in both cases the coefficients are of the opposite sign than what was expected.

The fourth column focuses on the games with an unexpected result. A match result is deemed unexpected when the result of the game is not the same as the result predicted ex-ante by the lowest betting odd. Similar to the subgroup consisting of important matches only, when isolating matches with unexpected result both the coefficients for wins and losses are negative. The same holds for local sponsors and finals, in which both the win and loss coefficients have the expected sign.

One issue that needs to be addressed at this point is the potential for the win and loss dummies in the subsequent regressions to be perfectly negatively correlated, which is commonly known as a problem of multicollinearity. Table B1 in appendix B presents a correlogram, indicating the correlation between the independent variables. It can be seen that there is no multicollinearity problem, as the correlation coefficient between wins and losses equals -0.66. Imdadullah et al. (2016) describe that multicollinearity exists when correlation coefficients are below -0.8.

B. Win/Loss effects

Table 2 reports the findings of the first set of regressions.

Table 2.

Regression outcomes for wins and losses with abnormal returns as the dependent variable, with wins and losses as the independent dummy variables.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The win and loss effect are represented by the estimates of β1WIN and β2LOSS. Regression outcomes were generated from the equation:

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡 + εi,t

*= Significant at the 10% level, **= Significant at the 5% level, ***= Significant at the 1% level

Wins

N α P-Value β1WIN P-Value

All games 508 0.0048 0.1144 -0.0044 0.3498

Losses

N α P-Value β2LOSS P-Value

All games 479 0.0017 0.5536 0.0031 0.5125

The output presented in table 2 is a product of regressions that were run on the complete sample.

The top column outlines the effect of wins in Champions League matches on the stock prices of the jersey sponsoring companies and the bottom column displays the effect of a loss. The beta coefficient in both the win and loss column, have the opposite sign of what was

(15)

Table 3 extends the analysis by distinguishing between the abnormal returns following group matches and the abnormal returns following knockout matches.

Table 3.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The outcomes of the regression analyses conducted in this table compare group matches to knockout matches. The win and loss effects are represented by the estimates of β1WIN and β2LOSS.

Regression outcomes were generated from the equations: 𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡 + εi,t

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐾𝑂𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐾𝑂𝑡 + εi,t

Wins

Group games 384 0.0057 0.1382 -0.0043 0.4766

Knockout games 124 0.0003 0.8597 -0.0018 0.5340

Losses

Group games 350 0.0021 0.5771 0.0051 0.4079

Knockout games 129 0.0012 0.5317 -0.0037 0.2021

The findings from table 3 also indicate no significant evidence yet to ascertain a relation between football results and stock returns for the jersey sponsors. When discerning wins and losses in group games from wins and losses in the more crucial knockout games none of the coefficients are significant. Wins in group as well as knockout games generated a

coefficient that is negative. Losses in a group game yielded a positive effect on the abnormal returns for the sponsors. Based on the findings presented in table 3, it cannot be suggested that wins and losses in group matches yield different results than wins and losses in

knockout matches as the results in both tables are not statistically distinguishable from zero.

(16)

Table 4.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The outcomes of the regression analyses conducted in this table compare

expected match results to unexpected match results. The win and loss effects are represented by the estimates of β1WIN and β2LOSS. Regression outcomes were generated from the equations:

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡 + εi,t

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡 + εi,t

Wins

Games with expected result 398 0.0084 0.1136 -0.0074 0.3152

Games with unexpected

result 111 0.0007 0.6575 -0.0012 0.7069

Losses

Games with expected result 344 0.0013 0.7857 0.0071 0.3383

Games with unexpected

result 136 0.0025 0.1051 -0.0077 ** 0.0104

As was mentioned before, a match outcome is considered unexpected when it does not reconcile with the outcome predicted by the lowest betting odd. The findings from table 4 clearly indicate that the match expectations are influential when losses are concerned. Similar to the findings in the previous two tables, wins do not yield significant abnormal returns for jersey sponsors, not even when the win was unexpected. Unexpected losses on the other hand, do impact stock prices as can be seen from the second column in table 4. A loss effect of -77 basis points is found. Notably, unexpected wins and unexpected losses by football clubs do not yield symmetrical abnormal returns for the sponsoring enterprises. Significant negative abnormal returns following unexpected losses are observed, whereas unexpected victories resulted in insignificant abnormal returns. This is in accordance with the allegiance bias phenomenon reported earlier and offers evidence to reject hypothesis 2, which states that the abnormal returns following a win do not differ from the abnormal returns following a loss.

(17)

Table 5.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The outcomes of the regression analyses conducted in this table compare

unexpected results in group matches to unexpected results in knockout matches. The win and loss effects are represented by the estimates of β1WIN and β2LOSS. Regression outcomes were generated from the equations:

𝐴𝑅𝑖,𝑡 α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡 + β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐺𝑅𝑂𝑈𝑃𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡+ εi,t

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐾𝑂𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡 + β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐾𝑂𝑡∗ 𝑈𝑁𝐸𝑋𝑃𝐸𝐶𝑇𝐸𝐷𝑡+ εi,t

Wins

Group games with unexpected result 83 0.0009 0.6193 -0.0035 0.3392 Knockout games with unexpected result 28 0.0000 0.9924 0.0060 0.3723

Losses

Group games with unexpected result 91 0.0012 0.4834 -0.0047 0.1849 Knockout games with unexpected result 45 0.0087 0.0198 -0.0172 *** 0.0032

The main aim of the regression results depicted in table 5 is to re-evaluate hypothesis 3, which states that there is no difference in results between group matches and knockout matches. When discerning group games from knockout games, there was no evidence to reject this hypothesis. However, when the primary focus is on matches with unexpected results there is evidence suggesting a difference in significance between group and knockout matches. Similar to the results obtained thus far, unexpected wins do not yield any

significant abnormal returns and neither do unexpected losses in group games. Losses in knockout matches on the other hand, do lead to negative abnormal returns of -172 basis points. This effect is larger in magnitude than the loss effect documented in table 4, which shows that game importance is to be considered relevant.

(18)

Table 6.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The outcomes of the regression analyses conducted in this table researches a subsample of local sponsors. The win and loss effects are represented by the estimates of β1WIN and β2LOSS. Regression outcomes were generated from the equation:

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐿𝑂𝐶𝐴𝐿𝑖+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐿𝑂𝐶𝐴𝐿𝑖 + εi,t

Wins

Local sponsors 245 0.0003 0.7817 0.0003 0.8254

Losses

Local sponsors 246 0.0010 0.3117 -0.0014 0.3515

The results from table 6 reveal that specifically examining abnormal returns from local sponsors does not yield any significant outcomes. When comparing these results to those found in table 2 it becomes apparent that the win and loss coefficients have the expected signs when local companies are examined solely, as opposed to when local and non-local companies are examined simultaneously. However, the coefficients are not statistically different from zero. Hence, there is no support to reject hypothesis 5, which states that the results do not change when local sponsors are examined separately.

Finally, table 7 presents the results of the regressions which isolate the finals in the sample.

Table 7.

The outcomes presented in this table are from the full sample of abnormal returns from all 54 companies in the sample and all 1236 events. The outcomes of the regression analyses conducted in this table isolate the effect of only the finals in the sample. The win and loss effects are represented by the estimates of β1WIN and β2LOSS. Regression outcomes were generated from the equation:

𝐴𝑅𝑖,𝑡= α + β1𝑊𝐼𝑁𝑖,𝑡∗ 𝐹𝐼𝑁𝐴𝐿𝑡+ β2𝐿𝑂𝑆𝑆𝑖,𝑡∗ 𝐹𝐼𝑁𝐴𝐿𝑡 +εi,t

Wins

Finals 3 -0.0016 0.7386 0.0179 * 0.0804

Losses

(19)

The results in table 7 exhibit the abnormal returns generated on the first trading day after the finals. It can be seen that slight positive abnormal returns are recorded after a win and slight negative abnormal returns are found after a loss. Both coefficient amount to 179 basis points but are only weakly significant. On top of the moderate significance of the win and loss coefficients, their explanatory power is also low due to the very limited amount of observations (11). What is more, the coefficient capturing the effect of a win in the final is measured over only three observations. Therefore, table 7 offers mild evidence in support of rejecting hypothesis 6. Yet, no strong inferences can be drawn. Moreover, since draws are not a possibility in a final, the multicollinearity problem arises in this particular analysis. This presents more reasons to not draw too strong inferences from table 7 altogether.

Overall, football results only significantly influence stock returns in the case of an

unexpected loss. Every football match that ended with a win or any other result that was forecasted by the odds ex-ante did not yield any significant abnormal returns, regardless of the match’s importance.

Moreover, when separating group matches from knockout matches, a difference in

significance was observed for the unexpected losses. An unexpected loss in the group stage of the tournament does not lead to a significant drop in the jersey sponsor’s stock price, whereas an unexpected loss in the knockout stage does.

The results found do not become more pronounced when local sponsors are examined in isolation. Unexpected losses in knockout matches are found to have the most significant impact on jersey sponsor’s stock prices.

Subsequently, the aforementioned findings will be analysed with regard to the economic and behavioural mechanism in the next part of this section.

C. Discussion

The outcomes of the event study and subsequent regression analyses provide evidence regarding the hypotheses that were outlined in section II of this thesis. The first and main hypothesis cannot be rejected on the basis of the findings in table 1 for the entire sample. However, as the previous section illustrated, football results have a significant effect on the stock price of jersey sponsors when pre-game match expectations are considered. Notably, losses in the Champions League that were unexpected yield significant negative abnormal returns of 77 basis points for the companies that appear on the participants’ jerseys. The magnitude of the loss effects can be placed in between the loss effect documented by Renneboog & Vanbrabant and Hanke & Kirchler (2013), who report loss effects of 140 and -30 basis points. Edmans et al. (2007) on the other hand, report a considerably larger loss effect at -700 basis points.

(20)

Game importance is found to be relevant in determining the magnitude of the loss effect. Although, solely discerning all group games from all knockout games did not generate any significant abnormal returns, a significant difference of 125 basis points is found when separating unexpected results in group matches from unexpected results in knockout matches. Hence, hypothesis 3 can also be rejected, which is in line with the results found by Renneboog & Vanbrabant (2002), Edmans et al. (2007), Hanke & Kirchler (2013) and

Eisdorfer & Kohl (2015). Rejecting the third hypothesis is in favour of the economic

explanation for the existence of significant abnormal returns. The fact that the loss effect is of a larger magnitude for knockout games implies that investors recognize the foregone economic benefits of another 90 to 180 minutes of exposure upon elimination.

As was mentioned at the beginning of this section, ex-ante match outcome expectations are the crucial factor that separates the significant losses from the insignificant losses.

Also, the analysis displays that the difference between expected and unexpected losses aggregates to 148 basis points, which offers strong support to reject hypothesis 4. This result is in line with the findings by Edmans et al. (2007) and Hanke & Kirchler (2013) and

corresponds to a semi-efficient market in which objective win and loss probabilities are already incorporated into stock prices. Since results that were already expected ex-ante do not have any financial implications for the jersey sponsor as opposed to unexpected results, it can be argued that investors act rationally and influence stock prices through the

economical mechanism.

Furthermore, this research ascertains that the results that were previously discussed are not amplified when the analyses is performed over a subsample consisting exclusively of local jersey sponsors. Both the coefficients for wins and losses are not statistically different from zero, as was the case when examining the full sample. Hypothesis 5, hence, cannot be rejected. Edmans et al. (2007) found a significant larger effect for small cap stocks compared to large cap stocks, which they attributed to sentiment trading. The same intuition applies behind the reason for separating local sponsors from the full sample, as local sponsors are more likely to be financed by local investors with a stronger affinity towards a Champions League participant in their home country. However, since the subset of local sponsors did not generate more pronounced results, no further support is provided for the behavioural mechanism, through sentiment trading, in this particular research.

Finally, it was difficult to determine whether to reject or not to reject hypothesis 6. As mentioned previously, there is a very limited amount of observations available for the finals. The results hint at some evidence in favour of the behavioural mechanism, but the results are influenced by multicollinearity between the win and loss variables.

D. Statistical robustness check

Prior to conducting the statistical robustness check, it was tested whether the residuals follow the pattern of a normal distribution. Figure B1 in appendix B presents a histogram with the corresponding Jarque-Bera statistic, revealing a normal distribution. Notably, this histogram depicts the distribution of the residuals from table 5, where the most significant result is established.

(21)

eliminating the 10% largest and 10% lowest observations, similar to what Edmans et al. (2007) did. Appendix C presents a table with all the analyses that were performed in the beginning of this section, conducted on a trimmed sample without outliers. The results from the table indicate that the results are not completely insensitive to outliers. However, the general implications derived from the analysis still hold. Looking at victories first, the elimination of outliers yielded a slight positive effect for wins that were unexpected, which was not recorded in the main analysis. On the other hand a slightly negative reaction is reported following victories by teams with a local sponsor. Most remarkably is that expected victories are found to yield highly significant negative abnormal returns of -26 basis points. Concerning the findings of this thesis, these results only further exacerbate the absence of an effect after a win in the Champions League on jersey sponsors’ stock prices.

Likewise, the unexpected loss effect is robust to the elimination of outliers. Unexpected losses remain significantly negatively related to jersey sponsors’ stock prices and the effect aggravates when looking at unexpected losses in knockout rounds. Despite a slight change in the coefficients, these effects remain consistent in significance. Another interesting finding is that a loss in knockout matches overall (expected and unexpected) lead to a stock price decrease of 29 basis points, an effect that was not found in the main analysis. This only further proves the difference in results between group matches and knockout matches that was already established in the main analysis of this report.

VI. Conclusion

A number of papers have established a connection between football results and asset prices. Motivated by that research, this thesis investigates whether football results can influence the stock price of the enterprises that are jersey sponsors of a football team. This thesis documents a significant loss effect of -77 basis points, when that loss was unexpected. Moreover, when the unanticipated loss occurred in a knockout match, the effect

exacerbated to -172 points. There is no corresponding effect for wins. All these findings are robust against omitting outliers in the sample.

Altogether, the findings in this paper lead to the belief that the loss effect can be attributed to an economical mechanism, in which investors respond to relevant cash flow information about the jersey sponsors. This explanation is favored over a behavioral explanation of the abnormal returns for multiple reasons. Firstly, unexpected losses in knockout matches, which are believed in this research to have strong economic implications for jersey sponsors, generate stronger results than unexpected losses in group matches. Secondly, unexpected losses yield larger abnormal returns than expected losses. Thirdly, the results do not amplify, but rather weaken, when a subsample of local sponsors is examined. The intuition is that local sponsors are specifically financed by local investors with stronger emotional ties to the football team, which would induce sentiment trading.

(22)

corporations to thoroughly assess the potential of the team they consider to sponsor in achieving their expectations in the Champions League.

As a final remark, several limitations can be addressed with respect to this thesis, which offer implications for further research. Notably, when conducting the event study, only two broad market indices from the US and Europe were used as benchmarks. However, to capture abnormal returns from companies that are not US or Europe based, more specific indices would perhaps appear to be a better benchmark. Utilization of these indices would be possible for those researchers that would have such indices at their availability. Also, the event window might not have perfectly considered time zone differences between

(23)

References

- Ashton, John K., Gerrard, B., Hudson, R., 2003. Economic impact of national sporting success: evidence from the London stock exchange. Applied Economics Letters 10.12, 783-785.

- Baker, M., Wurgler, J., 2006. Investor sentiment and the cross‐section of stock returns. The Journal of Finance 61.4, 1645-1680.

-

Burton G., Fama E., F., 1970. Efficient capital markets: A review of theory and empirical work. The Journal of Finance 25.2, 383-417.

- Edmans, A., Garcia, D. and Norli, O. 2007. Sports sentiment and stock returns. Journal of Finance 62(4), 1967-1998.

- Eisdorfer, A., Kohl, E., 2015. Corporate Sport Sponsorship and Stock Returns: Evidence from the NFL. Unpublished working paper. University of Connecticut. - French, K., R., Poterba, J., M., 1991. Investor diversification and international equity

markets. National Bureau of Economic Research, No. w3609.

- Hanke, M., Kirchler, M., 2013. Football championships and jersey sponsors’ stock prices: an empirical investigation. The European Journal of Finance 19.3, 228-241.

-

Hirshleifer, D., Shumway, T., 2003. Good day sunshine: Stock returns and the weather. The Journal of Finance 58.3, 1009-1032.

-

Imdadullah, M., Aslam, M., Altaf, S., 2016. Mctest: An R Package for Detection of Collinearity among Regressors. R JOURNAL 8.2, 495-505.

-

MacKinlay, A. C., 1997. Event studies in economics and finance. Journal of economic literature 35.1, 13-39.

(24)

Appendix

A. List of companies

Table A1.

Complete list of companies included in the analysis.

Sponsor Location of Headquarters Sponsored Football Club(s)Location of Football Club Area of business

888.com Gibraltar, Spain Sevilla Sevilla, Spain Betting

Aegon Den Haag, The Netherlands Ajax Amsterdam, The Netherlands Financial service AIA Hongkong Tottenham Hotspur London, United Kingdom Insurance AIG New York, United States Manchester United Manchester, United Kingdom Insurance Aon London, United Kingdom Manchester United Manchester, United Kingdom Consultancy Arcelormittal Luxemburg, Luxemburg Oțelul Galați Galați, Romania Construction Autonomy (Hewlett-Packard)* Palo Alto, United States Tottenham Hotspur London, United Kingdom IT

Axa Paris, France Braga Braga, Portugal Insurance

BNP Paribas Paris, France Anderlecht Brussels, Belgium Banking Bwin (GVC Holdings)* Isle of Man AC Milan Milan, Italy Betting Bwin (GVC Holdings)* Isle of Man Real Madrid Madrid, Spain

Carling (Molson)* Denver, United States Celtic Glasgow, Scotland Spirits Carling (Molson)* Denver, United States Rangers Glasgow, Scotland

Carlsberg Copenhagen, Denmark Liverpool Liverpool, England Spirits Carlsberg Copenhagen, Denmark FC Copenhagen Copenhagen, Denmark

Chevrolet (General Motors)* Detroit, United States Manchester United Manchester, United Kingdom Automobiles Citibank New York, United States Werder Bremen Bremen, Germany Banking Citibank New York, United States FCSB Bucharest, Romania

Cola Turka (Ulker)* Istanbul, Turkey Besiktas Istanbul, Turkey Food Daikin Osaka, Japan Club Brugge Bruges, Belgium Electronics Deutsche Telekom Bonn, Germany Bayern München Munich, Germany Telecommunication Doosan Group Seoul, South Korea Viktoria Plzen Plzen, Czech Republic Construction Enbw Karlsruhe, Germany VFB Stuttgart Stuttgart, Germany Energy Evonik Essen, Germany Borussia Dortmund Dortmund, Germany Chemicals Fiat Chrysler Amsterdam, The Netherlands Juventus Turin, Italy Automobiles Fortuna entertainment group Amsterdam, The Netherlands Legia Warszawa Warsaw, Poland Betting Gazprom Moscow, Russia Zenit St. Petersburg St. Petersburg, Russia Natural gas Generali Trieste, Italy Standard Liège Liège, Belgium Insurance Honda Tokyo, Japan Maccabi Haifa Haifa, Israel Automobiles Hyundai Seoul, South Korea Olympique Lyon Lyon, France Automobiles KIA Seoul, South Korea Bordeaux Bordeaux, France Automobiles KIA Seoul, South Korea Atletico Madrid Madrid, Spain

LG Seoul, South Korea Bayer Leverkusen Leverkusen, Germany Electronics Lukoil Moscow, Russia Spartak Moscow Moscow, Russia Oil NOS Lisbon, Portugal Sporting Lisbon Lisbon, Portugal Media Novartis Basel, Switzerland FC Basel Basel, Switzerland Healthcare Novotel (Accor)* Île-de-France, France Olympique Lyon Lyon, France Hospitality

Partouche Paris, France Lille Lille, France Hospitality

Philips Amsterdam, The Netherlands PSV Eindhoven, The Netherlands Electronics Plus 500 Israel Atletico Madrid Madrid, Spain Financial service Portugal Telecom Lisbon, Portugal Benfica Lisbon, Portugal Telecommunication Portugal Telecom Lisbon, Portugal FC Porto Porto, Portugal

Portugal Telecom Lisbon, Portugal Sporting Lisbon Lisbon, Portugal

Rosseti Moscow, Russia CSKA Moscow Moscow, Russia Electricity Samsung Seoul, South Korea Chelsea London, United Kingdom Electronics

Solar Jinko China Valencia Valencia, Spain Energy

Spar Nord Aalborg, Denmark Aalborg Aalborg, Denmark Banking Standard Chartered London, United Kingdom Liverpool Liverpool, England Banking Sunpower San Jose, United States Bayer Leverkusen Leverkusen, Germany Energy Tennents (C&C Group)* Dublin, Ireland Celtic Glasgow, Scotland Spirits Tennents (C&C Group)* Dublin, Ireland Rangers Glasgow, Scotland

Toyota Toyota, Japan Fiorentina Florence, Italy Automobiles Toyota Toyota, Japan Valencia Valencia, Spain

TUI Hannover, Germany FC Twente Enschede, The Netherlands Tourism TUI Hannover, Germany FC Zürich Zürich, Switzerland

Türk Telecom Altındağ, Turkey Fenerbahce Istanbul, Turkey Telecommunication Türk Telecom Altındağ, Turkey Galatasaray Istanbul, Turkey

Türk Telecom Altındağ, Turkey Trabzonspor Trabzon, Turkey

Turkcell Maltepe, Turkey Bursaspor Bursa, Turkey Telecommunication Turkish Airlines Istanbul, Turkey Galatasaray Istanbul, Turkey Aviation

Veolia Paris, France Olympique Lyon Lyon, France Environmental service Verbund AG Vienna, Austria Austria Wien Vienna, Austria Electricity

Vodafone London, United Kingdom Besiktas Istanbul, Turkey Telecommunication Vodafone London, United Kingdom Olympiakos Piraeus, Greece

(25)

B. Further descriptive statistics Table B1

Correlogram abnormal returns, win dummy and loss dummy

Figure B1

Distribution residuals for the unexpected results in knockout matches.

Abnormal Returns Win Loss

Abnormal Returns 1.0000 -0.0266 0.0186

Win -0.0266 1 -0.6634

(26)

C. Results robustness check Table C1

Outcomes of the robustness check: elimination of outliers.

Wins N α P-Value β1WIN P-Value All games 418 0.0000 0.9512 0.0006 0.3779 Group games 315 0.0000 0.9671 0.0002 0.7814 Knockout games 103 -0.0002 0.8439 0.0017 0.2113

Games with unexpected result 93 -0.0002 0.7242 0.0022 * 0.0845

Group games with unexpected result 72 -0.0003 0.7203 0.0015 0.2920

Knockout games with unexpected result 21 -0.0001 0.9406 0.0045 0.1054

All games local sponsors 206 0.0009 0.0790 -0.0013 * 0.0945

Finals 2 0.0019 0.5882 0.0070 0.3702 Losses N α P-Value β2LOSS P-Value All games 383 0.0004 0.2972 -0.0006 0.4084 Group games 280 0.0000 0.9299 0.0002 0.8168 Knockout games 103 0.0019 0.0434 -0.0029 ** 0.0369

Games with unexpected result 106 0.0010 0.1192 -0.0024 ** 0.0460

Group games with unexpected result 76 0.0004 0.5351 -0.0013 0.3628

Knockout games with unexpected result 30 0.0036 0.0196 -0.0065 *** 0.0094

All games local sponsors 191 0.0000 0.9825 0.0009 0.2619