MIDWEEK EFFECT ON SOCCER PERFORMANCE: EVIDENCE FROM THE GERMAN BUNDESLIGA.
Krumer, Alex ; Lechner, Michael
MIDWEEK EFFECT ON SOCCER PERFORMANCE: EVIDENCE FROM THE GERMAN BUNDESLIGA.
I. INTRODUCTION
The effect of the schedule on the performance of individuals or
groups in competitive environments has been widely investigated in the
recent behavioral economics literature (Cohen-Zada, Krumer, and
Shtudiner 2017; Page and Page 2010; Palacios-Huerta 2014). The increased
interest stems from a possible unfair ex-post advantage to one of the
contestants caused by psychological or strategic effects driven by the
schedule, which, ex ante, appeared to be fair. Therefore, fairness in
scheduling actions in competitive environments can be an important
economic issue. To be more specific, an unfair schedule may create
selection efficiency concerns by reducing a "better"
contestant's probability of winning. In addition, it can also harm
contestants' future revenues and therefore affect their willingness
to exert efforts in the present. Consequently, to maximize both,
selection efficiency and effort, it is important for contest designers
to minimize any possible advantage to one of the contestants that stems
from an unbalanced schedule.
The role of scheduling in tournament settings was discussed in
different types of contests. For example, Klumpp and Polborn (2006)
described the unfairness of the primary party-based election system used
in the United States to nominate a candidate of one of the major parties
for the participation in presidential elections. According to the
authors, this structure is unfair, since it shifts too much power to the
voters in the early primary states. Page and Page (2010) presented a
systematic bias in the sequential evaluation of performance, namely that
contestants who performed in the later serial position in the popular
Idol series had a significantly larger advantage with respect to a
positive evaluation by the jury. (1) De Bruin (2005) found similar
results for the Eurovision song contest and for the World and European
figure skating championships. Similarly, Glejser and Heyndels (2001)
found that piano finalists who performed later in the final week of the
prestigious Queen Elisabeth music contests obtained a higher rank.
However, this relationship was not statistically significant in the
violin competition. In another study, Page and Page (2007) showed that
there is advantage of playing in the second home leg game in soccer
European tournaments. Krumer (2013) explained this result theoretically,
assuming that the winner of a first stage has a psychological advantage
in the second stage. Finally, to insure the fairness of multistage
contests with a sequential order of moves, Palacios-Huerta (2012)
proposed to use the so-called Prouhet-Thue-Morse sequence, where the
sequence of the first n moves is the exact mirror image of the next n
moves.
The aim of this study is to evaluate the effect of midweek matches
on the home advantage in the German Bundesliga 1 (hereafter Bundesliga),
which is the highest division of German male soccer.2 The relevance of
this question stems from the fact that some teams play more midweek
matches at their home stadium than others. Therefore, a schedule is
considered as fair if ex ante all teams in the Bundesliga have the same
probability to convert the home advantage into success, given their
individual characteristics, regardless of the weekday.
In general, the home advantage phenomenon is a well-established
feature in sports' competitions. Courneya and Carron (1992) defined
the home advantage as "the consistent finding that home teams in
sports competitions win over 50% of the games played under a balanced
home and away schedule" (p. 13). In their book, Moskowitz and
Wertheim (2011) survey 19 different sports leagues covering more than 40
countries between the years 1871 and 2009 and showed that the within
league home field advantage "is almost eerily constant through
time" (p. 113). The percentage of games won by the home teams in
these leagues varied between 53.3% and 69.1%. The home advantage
phenomenon can be attributed to crowd noise (Pettersson-Lidbom and Priks
2010), positive psychological states during home games (Terry, Walrond,
and Carron 1998), familiarity with facilities (Pollard 2002), increased
level of testosterone in the players (Neave and Wolfson 2003), distance
between cities (Oberhofer, Philippovich, and Winner 2010), or referee
bias (Dohmen and Sauermann 2016; Garicano, Palacios-Huerta, and
Prendergast 2005; Sutter and Kocher 2004). Whatever the reason, it is
clear that a certain home advantage is a feature of many sporting
events.
Despite the fact that home advantage has been widely analyzed in
the literature, to the best of our knowledge, there is no study on the
effect of playing midweek on the size of the home advantage. It is,
however, well documented that midweek matches attract a lower crowd and
lower TV ratings in English soccer (Buraimo 2008; Buraimo and Simmons
2015; Forrest and Simmons 2006), U.S. Major League Baseball (Knowles,
Sherony, and Haupert 1992), English cricket (Schofield 1983), and the
North American National Hockey League (Paul 2003). In a recent study,
Wang, Goossens, and Vandebroek (2016) showed that stadium fans in
Belgium's soccer Pro Legue show a very high aversion to matches
that take place on Wednesdays. Therefore, combining the literatures on
the effect of crowds on home advantage and on the relationship between
midweek games and the size of the crowd, suggests that the home
advantage may differ between weekend and midweek days.
Indeed, based on 2,013 Bundesliga matches in the seasons 2007-2008
to 2016-2017, we find that the home advantage in the midweek matches
completely disappears. According to our propensity score matching
analysis below, playing in the midweek reduces the difference in points
between home and away teams from 0.48 points over the weekend to
essentially zero. (3) In our robustness checks, we rule out other
possible explanations like fatigue, the size of the crowd, or the effect
of a specific weekend day. (4) Interestingly, we also find that both
teams commit less fouls in the midweek matches. However, this reduction
is significantly higher for home teams than for away teams. Therefore,
it appears unlikely that referee bias in favor of away teams occurring
in midweek matches explains the midweek effect. Another possible
explanation of our results can be linked to the decreased importance of
the midweek matches as perceived by the home teams' players. This
may reduce the association of testosterone, which is known to enhance
performance and aggressiveness, and territoriality, as described by
Neave and Wolfson (2003). (5)
To illustrate a possible relationship between an unbalanced
schedule and the resulting monetary rewards one can look at the
distribution of revenues from TV broadcast contracts in the season
2015-2016. For example, the fourth best team with regard to TV revenues
in the Bundesliga 2, SC Paderborn 07, which was relegated from the
Bundesliga in the 2014-2015 season, received 9.89 million Euros. In
stark contrast, the lowest TV revenues in the Bundesliga were equal to
20.19 million Euros and were earned by SV Darmstadt 98. This means that
if SC Paderborn 07 had stayed in the Bundesliga, its revenue from TV
alone would have been at least 10.3 million Euros higher (not counting
all other revenues from ticketing, advertising, and so on).6 The case of
SC Paderborn 07 is interesting also for the reason that in the
relegation season, before the last round, this team had one point less
than the Hamburger SV, a team that finally survived in the Bundesliga.
In that season, each team was scheduled to play three matches in the
midweek rounds. SC Paderborn 07 played two midweek matches at home. One
of these games was against Hamburger SV in which the latter won.
However, Hamburger SV had to play only one midweek match at home. Of
course, we do not claim that the main reason for SC Paderborn 07's
relegation was the fact that this team played at home in the midweek
round against its closest rival in the relegation fight. However, it is
clear that the teams had different incentives for winning their last
game, which, theoretically, leads to different winning probabilities
(Krumer, Megidish, and Sela 2017). Therefore, in a very tight league,
where every point is important, an uneven schedule may have an effect on
interim rankings and as a result on the final rankings as well (and thus
on future revenues).
The remainder of the study is organized as follows: Section II
describes the Bundesliga schedule. The data and some descriptive results
are presented in Section III. Section IV presents the estimation
strategy. The results are contained in Section V. Finally, in Section VI
we offer concluding remarks.
II. DESCRIPTION OF THE BUNDESLIGA SCHEDULE
The German Bundesliga consists of 18 clubs. The matches are
organized as a round-robin system. Each round consists of nine different
matches, such that in each round, each team plays only once. (7) In the
first half of the season (typically August to December), every team
plays in 17 different matches against all other teams. Half of the
matches are at home. In the second half of the season (January to May),
each team plays again against all other teams, but teams played at home
in the first half of the season are now playing away. Thus, in total,
each team plays 34 matches, 17 of them at home, and another 17 away from
home. There is a break of about 1 month between the first and the second
half of a season without any matches.
The matches usually take place on weekends. A typical weekend round
consists of one match on Friday at 20:30 hours, several matches on
Saturday, usually at 15:30hours and one game at 18:30 hours. The
remaining matches take place on Sundays, usually at 15:30 and 17:30
hours. However, there are some rounds that take place during the midweek
days. These matches always take place on Tuesdays and Wednesdays at
20:00 hours. The reason for the existence of these midweek rounds is
that the schedule should take into account the winter break (weather
conditions), the summer break (players' vacation and recovery
time), as well as the international tournaments, such as the FIFA World
Cup and UEFA European Championship that take place every 2 years in June
and July. Therefore, as there are not enough weekends between August and
May, several rounds are played in midweek. In addition, a few regular
weekend matches that were postponed due to weather conditions also take
place in midweek. (8)
At the end of a season, the final table determines which teams
participate in the following season's European club tournaments,
such as the Champions League, which is the most prestigious club
tournament in Europe, and the Europe League (former UEFA Cup), which
also yields big monetary rewards. For example, in the 2015-2016 season,
the four highest ranked teams participated in the Champions League (the
fourth ranked team had to play an additional international qualification
game to participate in the group stage of the Champions League). Teams
in the fifth to seventh position played in the Europe League (this may
depend also on the outcome of an elimination tournament, called the
"DFB-Pokal"). In addition, the two worse ranked clubs relegate
to the lower division and the team ranked 16th has to participate in the
relegation play-offs against the team that was ranked third in the
Bundesliga 2 for the right to play in the Bundesliga in the following
year. (9) Note that the different outcomes have financial consequences
for the clubs in the next season.
III. DATA AND DESCRIPTIVE RESULTS
A. Data Base
We collected data on all the matches in the Bundesliga from the
2006-2007 season up to the 10th round of the 2016-2017 season (latest
data available when revising this study). The data collection starts
with the 2006-2007 season, because a large amount of important
information was unavailable prior to that season. In total, these data
cover 3,150 matches. However, data on 306 matches played in the
2006-2007 season were used to control for the previous season's
team's characteristics. In addition, we disregard matches of teams
playing the European competitions just before and after these games
(UEFA Champions League or UEFA Europa League), because such matches may
create different allocation of efforts (for example, saving best players
to more important European Cups matches, fatigue, or psychological
momentum). We also did not take into account matches in which a home
team did not play at its home stadium (e.g., Bayer Leverkusen in season
2008-2009), matches that took place after an international break in
which national teams played friendly or qualification games, and matches
in which one of the teams was already relegated or already won the
title. After dropping these games, 2,013 matches enter the estimation
(1,861 weekend games and 152 midweek games).
For every match, there is information on the exact day and hour,
attendance, and the final score. For each team in a particular game, we
observe the number of shots, shots on target (i.e., all shots that would
score in the absence of a goalkeeper), number of fouls, as well as the
number of corners and yellow and red cards. In addition, we use data
from the Transfermarkt website (www .transfermarkt.com) to proxy the
market value of each player of each team in every season. Finally, we
have data on the dates of the beginning and the end of work of each
coach, as well as data on the capacity of each stadium and on regional
economic characteristics. The data were collected from several web sites
(see Appendix D for the full list). (10)
B. Descriptive Statistics
To estimate the possible effect of playing midweek, we have a set
of possible outcome variables on the level of a single match between
home and away teams. The first two are defined as the number of points
obtained by home and away teams. In Table 1, we can see that in line
with the home advantage phenomenon, a home team attains on average 1.60
points per weekend match, which is significantly higher than the 1.33
points to be expected in the absence of a home advantage. (11) An away
team achieves on average 1.14 points per such match. However, in the
midweek matches we observe that the home advantage completely
disappears.
Next, we define the number of goals scored by home and away teams.
Based on these two variables, we also calculate the difference of goals
between home and away teams. Table 1 presents, not surprisingly, that on
average a home team scores more goals than an away team. Other possible
outcome variables are related to the shots and shots on target. Table 1
shows that home teams have higher values for both shots-related
variables.
Another dimension, which is interesting to investigate, is related
to the aggressiveness of the teams, such as committed fouls, and yellow,
and red cards. From the descriptive statistics, we learn that in general
home teams commit less fouls, which is probably translated to the lower
number of yellow and red cards.
C. Variables
To estimate the effect of midweek matches on performance, we code a
dummy variable that is one if a match was played in the midweek and zero
otherwise. We also use a rich set of variables that characterizes team
value and players' ability, game attendance, and the international
and national schedule (and the resulting demands on the international
players). In the following, we present some of the most important
measures (a more comprehensive list of variables appears in Appendix A).
To approximate teams' abilities, we use teams' monetary
values obtained from a popular soccer website, Transfermarkt, which is a
reliable data source that provides data on players' market values.
As reported by Bryson, Frick, and Simmons (2013), the coverage of
Transfermarkt is quite "impressive with information on 190,000
players across 330 football competitions" (p. 611). According to
Franck and Nuesch (2012), players' values are estimated by industry
experts and take into account salaries, signing fees, bonuses, and
transfer fees. The authors found that the correlation between values
evaluated by Transfermarkt and Kicker, another highly respected sport
magazine in Germany, is as high as 0.89. In addition, Frick (2006) found
that the correlation between salary information published by Kicker and
actual salaries for two seasons in Bundesliga is 0.8.
The players' values are used to create some additional
measures like the distribution of values between and within teams. More
specifically, for each team we compute the standard deviation of
players' values, the Herfindahl-Hirschman Index (HHI), which is
defined as the sum of the squares of the values shares of each player
within the team. In addition, we create some other
within-team-inequality related variables such as the ratio of different
players' values according to their ranking order in the team. For
example, one measure is the ratio between the top three players to
players ranked 9 to 11 according to their values within a team (see
Appendices A and B for more details). (12) It is important to note that
the goal of our empirical analysis is to evaluate the effect of midweek
matches rather than to determine players' values. We use
values-related measures only as covariates, which are supposed to
reflect teams' abilities. The teams' values measure strongly
correlates with teams' performance, suggesting that we measure
teams' abilities quite well. (13) In addition to players'
values, we also use several other variables that may reflect the level
of ability such as a dummy variable for a team's first season in
the Bundesliga after being promoted from Bundesliga 2, whether a team
dismissed its coach during a season, and teams' previous
season's characteristics, such as shots, corners, and yellow and
red cards. (14)
As found previously in the literature, the home advantage is
affected by the attendance level. Therefore, we create a measure to
reflect the attendance in a match. Our preferred measure, attendance as
share of the capacity of the stadium, is the ratio between the number of
viewers in a match and the maximal possible capacity of the respective
stadium. Table 1 demonstrates that this measure is lower in midweek than
in weekend matches. There is also information about the distance between
cities and public transport commuting time between cities.
In addition, we obtain information on other schedule-related
variables in international competitions such as two pre- and post-World
Cup and European Championships months, as well as the months in which
the Africa Cups of Nations took place. Furthermore, we take into account
different parts of the season, such that the beginning of the season is
defined as the first 11 rounds, the middle of the season is defined as
rounds 12 to 22, and the end of the season includes rounds 23 to 34. In
addition, we split the beginning of the season into several parts.
IV. ECONOMETRICS
A. The Causal Question
We are interested in learning the effect of playing in the middle
of the week on the success of the home team in terms of the variables
measuring different aspects of the outcome of a Bundesliga soccer game,
as described in the previous section. If the allocation of the midweek
games over the season were entirely random, then we would compare the
means of these variables for midweek matches to the means obtained for
weekend matches. The difference would be a consistent estimate of the
desired effect. However, scheduling is only partially random, since many
other considerations are considered when fixing a league schedule, like
weather, players' rest period, European Cups, National teams'
tournaments, among others. Furthermore, the distribution of the
characteristics shown in Table 1 already point to (small) deviations
from randomness. Such deviations need to be taken into account in any
estimation strategy, if they are correlated with the outcomes of
interest (e.g., Imbens and Wooldridge 2009), which are measures of the
success of the home team in our case. (15) Here, the data base available
is rich enough in terms of game and team characteristics that we opt for
a selection-on-observable strategy to identify the causal effects of
interest. As described previously, since the schedule of the Bundesliga
has to take into account winter and summer breaks, as well as
international tournaments, we expect to capture other schedule-related
characteristics, such as the periods of time that are associated with
the World Cups, European Championships, and the Africa Cup of Nations.
In addition, as suggested by the previous literature on the linkage
between midweek matches and the crowd, we use attendance-related
variables as well. Finally, we capture other differences related to
team, location, and timing by the variables described above.
B. Estimator Used
Since the previous section suggests that controlling for observable
characteristics will be sufficient to identify a causal effect, we face
two challenges. The first challenge is that we expect that the effect of
midweek games may be different for different clubs (e.g., depending on
their actual position in the season) and games (early vs. late in the
season) in the Bundesliga. Since the exact kind of heterogeneity is
unknown, and since a very flexible way of controlling for the various
confounding factors appears to be called for, we use a statistical
matching approach. To be more specific, we employ the
radius-matching-on-the-propensity-score matching estimator with bias
adjustment as suggested by Lechner, Miquel, and Wunsch (2011) because it
showed its superior finite sample and robustness properties in the
large-scale empirical Monte Carlo study conducted by Huber, Lechner, and
Wunsch (2013).
The second issue we face is how to exactly specify the propensity
score, which is the probability of a weekday match given the relevant
characteristics. The problem is that although we have prior knowledge on
the kind of variables needed, these considerations are uninformative
about exactly which measurements to use (like which functional form,
which interactions, or which particular measure of distance, like travel
time or kilometers, to mention just a few). In the past, researchers
used more or less ad hoc specifications that pass certain checks with
respect to the performance of the matching procedures, like so-called
balancing tests. However, recent advances in machine learning techniques
suggest using more principled variable selection procedures. In
particular, we employ the ideas of Belloni, Chernozhukov, and Hansen
(2014) of using the Least Absolute Shrinkage and Selection Operator
(LASSO). In fact we use the so-called Adaptive LASSO (Buhlmann and van
de Geer 2011; Hastie, Tibshirani, and Friedman 2009; Zou 2006) twice to
obtain the covariates. LASSO is a statistical procedure suggested by
Tibshirani (1996). It reduces the dimension of the model in some optimal
sense by adding a specific penalty term to the objective function of the
regression-type (linear, logistic, etc.) estimator. The LASSO penalty is
the sum of the absolute values of the regression coefficients. This
penalty term leads to the fact that many coefficients will be set to
zero. Therefore, if the true number of variables belonging to the model
is not too large (so-called sparsity assumption), the LASSO estimator
has good variable selection properties, and can thus be used to select
the variables to be included in the propensity score. A drawback of the
LASSO is that its properties deteriorate once the sparsity assumption
does not hold. However, in our case this should not be very important
since we are not really interested in exactly which variables are
included in the propensity score (which is however helpful for its
interpretation), but first of all aim at a good prediction of the score,
which, intuitively speaking, is sufficient to remove any selection bias
due to (observable) omitted variables.
Following Belloni, Chernozhukov, and Hansen (2014), there are two
LASSO estimations. The first LASSO estimation concerns the selection
equation. However, Belloni, Chernozhukov, and Hansen (2014) point out
that using just the selection equation for variable selection may not be
sufficient, as it might ignore variables that are mildly related to the
treatment (playing midweek), but heavily related to the outcome, and
thus should be controlled for. Thus, the second LASSO estimation has the
purpose to identify variables that are highly correlated with the
outcome (ignoring the midweek variable). The variables used in the
matching estimation are the union of variables selected by the LASSO in
either of those two steps. Although, there practical applications of
these new approaches are so far rare, they provide significant
improvements to the ad hoc variable selection procedures used so far.
The inference for the matching estimator is based on the weighted
bootstrap (see also the empirical Monte Carlo results on the performance
of different inference procedures investigated by Bodory et al. 2016)
ignoring the variable selection step (which is justified by the
LASSO's oracle properties).
V. RESULTS
Although the purpose of the propensity score estimation is only a
technical one, namely to allow the easy purging of the results from
selection effects, it is nevertheless interesting to see which variables
drive selection. Generally, as already apparent from Table 1, selection
effects are limited. They are substantially driven by the lower
attendance at midweek games as well as timing effects. The detailed
results can be found in Appendix B.
Table 2 shows the key results of this study, namely the effect of
playing midweek compared to playing weekends on various outcome
variables that may be used to characterize the results of soccer games.
The most important one is of course the expected number of points earned
(top of table): Playing midweek leads to an effect of about 0.64 points
in total, resulting from the home team losing about 0.29 points, while
the away team gains about 0.35 points (the asymmetry results from the
"3 points rule"). Considering the levels of expected points,
it becomes clear that the home team loses all its home advantages in
midweek games. In midweek games, the home and the away team (with the
same characteristics as the home team) can expect to earn about 1.33 and
1.49 points on average, respectively, while on weekends the (same) home
and away teams earn about 1.62 and 1.14 points, respectively. Similar
results appear when considering related measures, namely the goals
scored and the teams' shots (although not always statistically
significant).
Interestingly, when considering one aspect of the playing style by
looking at fouls and cards, it turns out that both teams commit fewer
fouls in midweek than on weekends. This fact points to a reduced
aggressiveness of both teams in midweek games, which might be attributed
to the general lower attendance in these games (although the level of
attendance is flexibly controlled for).
In addition, we investigate whether the results are robust to
including the information about attendance as control variable. If
spectators know that midweek games have no or a reduced home advantage,
then this fact may reduce their inclination to visit the games in
midweek and thus this variable is endogenous and we expect a bias of the
result toward zero when controlling for attendance. However, this is
clearly not the case as can be seen in Table A3. In fact, the results
without conditioning are even somewhat smaller.
We also compare midweek games to games played on Sundays and
Fridays. We report the results in the last two columns of Table A3. As
we can see, playing on midweek relative to Sunday has a significant and
substantial effect of 0.72 points in favor of an away team. As
previously, we also find a significantly lower number of fouls committed
by a home team in midweek games.
Despite the fact that there are only 205 games played on Friday, we
still find that an away team obtains 0.192 points more on midweek
relative to Friday with a p value of 8%. As in other specifications, we
find again that there are significantly less fouls committed by the home
team in midweek games compared to Friday games. The remaining results
are in the same direction as in other specifications. However, some of
them are not significant at conventional levels due to a much lower
number of observations available for estimation.
Finally, the explanation of the midweek effect might be that the
teams are more tired because they have a reduced rest period (about half
compared to a weekend game without prior midweek game) and thus are less
aggressive which might reduce the advantage of the home team. To
investigate this issue, we pool midweek games together with weekend
games after a midweek game and compare them to weekend games without a
prior midweek game. When doing so (Table G1 in Appendix S1, Supporting
Information), the effect becomes much smaller and statistically
insignificant, which is essentially ruling out that hypothesis.
One possible explanation of our results may be linked to the size
of the crowd. Despite the fact that we control for many crowd-related
variables, the size of the crowd may have indirect effects on
players' performance. As discussed previously, a smaller crowd in
the midweek matches is a well-known phenomenon across different sports
in different countries, and is found in our data base as well.
Therefore, it is likely that home team's players anticipate lower
attendance before these matches, such that home team's players may
consider midweek matches less important, because fewer viewers will
monitor their actions. (16) If so, then our findings on the disappearing
home advantage and lower aggressiveness of home teams, as measured by
the lower number of its fouls, are in line with the literature on
testosterone, which is known to enhance performance and aggressiveness.
(17) The point is that the above described anticipation can affect the
relationship between testosterone and territoriality that was described
by Neave and Wolfson (2003). In fact, these authors provide direct
evidence that the level of testosterone among soccer players was
significantly higher before home matches than before away matches or
training sessions. No difference was found between away matches and
training sessions. In addition, they found that the increase in
testosterone was higher before matches that were perceived as more
important. Similarly, Mazur, Booth, and Dabbs (1992) showed that a
pregame increase in testosterone is less likely to occur in the event
that is regarded as less important. These findings are in line with Ward
(1998) who studied the effect of opening day matches in Major League
Baseball on home advantage. In general, these matches are perceived as
very important and described as "... more than 'just another
game' ..." and as a " ... highly ritualistic and festive
occasion ..." Ward (1998, p. 280). Therefore, players and fans in
these matches are expected to be more motivated and as a result the home
advantage should be greater. Indeed, the author found significantly
higher winning probabilities of home teams on these opening day matches
than in other matches during a regular season.
VI. CONCLUSION
The main motivation of this study is the potential effect of
scheduling on the performance of high-profile agents in a real
competitive environment for which good productivity data are available.
We find that in the German soccer Bundesliga, the home teams perform
significantly worse in midweek matches than in weekend matches. In these
games, the home advantage, which is important in many different sports,
disappears completely. Since the midweek matches are unevenly allocated
among teams, this finding implies that the actual schedule of the
Bundesliga favors teams with fewer home (more away) games in midweek,
which may be considered as an unfair advantage for those teams. Such
unfair scheduling plays a significant role in determining teams'
season outcomes. Thus, it may change their income by tens of million
Euros per year.
Our results on lower aggressiveness and decreased performance of
home teams during their midweek matches, in addition to the smaller
crowd during these games, are in line with a biological literature on
the relationship between testosterone, the importance of the event,
performance, and territoriality. According to this literature, the
midweek matches may be perceived by the home teams' players as less
important, which may lead to decreased motivation and as a result to a
disappearing home advantage.
Despite being driven by different factors, the day of the week
effect on performance was also found in completely different settings.
In their seminal paper, Gibbons and Hess (1981) showed that there was a
negative mean return of financial assets for Mondays and a positive
return for Fridays. More recently, Siganos, Vagenas-Nanos, and
Verwijmeren (2014) found a significant relationship between
Sunday's sentiment and Monday's stock market characteristics,
highlighting that the behavioral factors like mood, optimism, and
happiness are responsible for their finding. Therefore, since the day of
the week effect on performance appears to be a more general behavioral
feature than just the effect in soccer, it should be taken into account
when evaluating individuals' performance in various fields, such as
labor market contests, financial markets, or sports competitions.
In particular, it is worthwhile for the Bundesliga management (DFL)
to allocate the midweek matches evenly among teams. For example, if
there was a midweek round in the first half of the season, then there
should be a midweek round in the second half of the season and it should
involve the same teams (even if this means changing the order of games
in both halves of the season). This more balanced schedule increases the
fairness of the tournament.
Finally, our methodology could be applied to examine other
performance-related features such as the effect of different parts of
the season, which may include national holidays, or the effect of a
dismissed head coach in the Bundesliga. In addition, despite several
possible fundamental differences between different leagues, such as the
share of midweek matches, existence of a winter break, and relative
frequency of matches on different days, we call for further research to
investigate the effect of allocation of matches on performance in
different sports leagues.
ABBREVIATION
LASSO; Least Absolute Shrinkage and Selection Operator
APPENDIX A: DESCRIPTIVE STATISTICS
The following table contains descriptive statistics for all
variables (outcomes, treatment, and controls) relevant for this study.
TABLE A1
Descriptive Statistics (Sample Means)
Variables Weekend Games Midweek Games
(N = 1,861) (N = 152)
Game outcomes
Points--home team 1.60 1.36
Points--away team 1.14 1.46
Goals--home team 1.56 1.51
Goals--away team 1.23 1.39
Shots--home team 14.38 13.99
Shots--away team 11.93 11.82
Shots on target--home team 5.13 5.01
Shots on target--away team 4.21 4.41
Fouls--home team 15.91 14.72
Fouls--away team 17.02 16.34
Yellow cards--home team 1.65 1.70
Yellow cards--away team 1.95 1.86
Red cards--home team 0.07 0.09
Red cards--away team 0.10 0.10
Corners--home team 5.56 5.36
Corners--away team 4.42 4.14
Game characteristics
Attendance 40,795 40,478
Attendance as share of 0.91 0.86
capacity of stadium
Attendance as share of 0.35 0.25
capacity of stadium
if higher than 0.99
Ln of stadium capacity 10.54 10.51
Distance between the cities 366 372
of the teams (km)
Public transport commuting 194 200
time between the cities
(minutes)
Africa Cup of Nations months 0.12 0.06
Two months before World Cup or 0.12 0.18
European Championship
Two months after World Cup or 0.07 0.24
European Championship
Season after World Cup or 0.44 0.55
European Championship
Rounds 1-11 0.27 0.43
(beginning of a season)
Rounds 12-22 0.33 0.20
(middle of a season)
Rounds 23-34 (end of a season) 0.40 0.37
August or September 0.18 0.36
Rounds 1-6 0.17 0.31
Team characteristics
Value of home team 86.7 109.7
(Mil. [euro])
Value of away team 90.3 103.9
(Mil. [euro])
Standardized value of -0.19 -0.02
home team
Standardized value of -0.15 0.00
away team
First season after 0.16 0.15
promotion--home team
First season after 0.15 0.11
promotion--away team
Ratio of top 3 to ranked 9-11 2.34 2.18
players' values--home team
Ratio of top 3 to ranked 9-11 2.34 2.48
players' values--away team
HHI of players' values--home 0.06 0.06
team
HHI of players' values--away 0.06 0.06
team
New coach--home team 0.18 0.11
New coach--away team 0.18 0.14
Previous season shots--home 14.56 15.04
team
Previous season shots--away 11.94 12.02
team
Previous season shots on 5.60 6.11
target--home team
Previous season shots on 4.53 4.85
target--away team
Previous season corners--home 5.64 5.76
team
Previous season corners--away 4.38 4.41
team
Previous season fouls--home 16.27 15.90
team
Previous season fouls--away 17.46 17.17
team
Previous season yellow 1.64 1.59
cards--home team
Previous season yellow 1.97 1.94
cards--away team
Previous season red 0.08 0.07
cards--home team
Previous season red 0.11 0.11
cards--away team
Variables Sunday Games Friday Games
(N = 421) (N = 205)
Game outcomes
Points--home team 1.57 1.68
Points--away team 1.17 1.05
Goals--home team 1.58 1.51
Goals--away team 1.30 0.96
Shots--home team 14.61 14.37
Shots--away team 12.03 11.32
Shots on target--home team 5.14 5.12
Shots on target--away team 4.41 3.76
Fouls--home team 15.87 15.71
Fouls--away team 16.94 16.55
Yellow cards--home team 1.60 1.63
Yellow cards--away team 1.94 1.97
Red cards--home team 0.07 0.05
Red cards--away team 0.10 0.06
Corners--home team 5.56 5.70
Corners--away team 4.43 4.13
Game characteristics
Attendance 41,051 44,036
Attendance as share of 0.91 0.91
capacity of stadium
Attendance as share of 0.33 0.41
capacity of stadium
if higher than 0.99
Ln of stadium capacity 10.55 10.61
Distance between the cities 357 357
of the teams (km)
Public transport commuting 189 186
time between the cities
(minutes)
Africa Cup of Nations months 0.13 0.13
Two months before World Cup or 0.09 0.08
European Championship
Two months after World Cup or 0.09 0.09
European Championship
Season after World Cup or 0.43 0.45
European Championship
Rounds 1-11 0.32 0.32
(beginning of a season)
Rounds 12-22 0.33 0.37
(middle of a season)
Rounds 23-34 (end of a season) 0.36 0.31
August or September 0.22 0.21
Rounds 1-6 0.20 0.19
Team characteristics
Value of home team 91.0 98.3
(Mil. [euro])
Value of away team 95.0 93.4
(Mil. [euro])
Standardized value of -0.12 -0.09
home team
Standardized value of -0.07 -0.14
away team
First season after 0.14 0.15
promotion--home team
First season after 0.10 0.14
promotion--away team
Ratio of top 3 to ranked 9-11 2.33 2.38
players' values--home team
Ratio of top 3 to ranked 9-11 2.37 2.30
players' values--away team
HHI of players' values--home 0.06 0.06
team
HHI of players' values--away 0.06 0.06
team
New coach--home team 0.17 0.13
New coach--away team 0.15 0.19
Previous season shots--home 14.67 14.75
team
Previous season shots--away 12.01 11.77
team
Previous season shots on 5.69 5.73
target--home team
Previous season shots on 4.61 4.43
target--away team
Previous season corners--home 5.69 5.68
team
Previous season corners--away 4.39 4.32
team
Previous season fouls--home 16.22 16.01
team
Previous season fouls--away 17.39 17.26
team
Previous season yellow 1.64 1.60
cards--home team
Previous season yellow 1.97 1.96
cards--away team
Previous season red 0.08 0.07
cards--home team
Previous season red 0.11 0.10
cards--away team
Variables Overall
(N = 2,013)
Game outcomes
Points--home team 1.58
Points--away team 1.16
Goals--home team 1.55
Goals--away team 1.24
Shots--home team 14.35
Shots--away team 11.88
Shots on target--home team 5.13
Shots on target--away team 4.22
Fouls--home team 15.82
Fouls--away team 16.97
Yellow cards--home team 1.65
Yellow cards--away team 1.95
Red cards--home team 0.08
Red cards--away team 0.10
Corners--home team 5.54
Corners--away team 4.40
Game characteristics
Attendance 40,771
Attendance as share of 0.90
capacity of stadium
Attendance as share of 0.34
capacity of stadium
if higher than 0.99
Ln of stadium capacity 10.54
Distance between the cities 366
of the teams (km)
Public transport commuting 195
time between the cities
(minutes)
Africa Cup of Nations months 0.12
Two months before World Cup or 0.11
European Championship
Two months after World Cup or 0.08
European Championship 0.45
Season after World Cup or
European Championship 0.28
Rounds 1-11
(beginning of a season)
Rounds 12-22 0.32
(middle of a season)
Rounds 23-34 (end of a season) 0.39
August or September 0.20
Rounds 1-6 0.18
Team characteristics
Value of home team 88.4
(Mil. [euro])
Value of away team 91.3
(Mil. [euro])
Standardized value of -0.18
home team
Standardized value of -0.14
away team
First season after 0.16
promotion--home team
First season after 0.15
promotion--away team
Ratio of top 3 to ranked 9-11 2.33
players' values--home team
Ratio of top 3 to ranked 9-11 2.35
players' values--away team
HHI of players' values--home 0.06
team
HHI of players' values--away 0.06
team
New coach--home team 0.17
New coach--away team 0.17
Previous season shots--home 14.60
team
Previous season shots--away 11.95
team
Previous season shots on 5.64
target--home team
Previous season shots on 4.56
target--away team
Previous season corners--home 5.65
team
Previous season corners--away 4.38
team
Previous season fouls--home 16.25
team
Previous season fouls--away 17.43
team
Previous season yellow 1.64
cards--home team
Previous season yellow 1.97
cards--away team
Previous season red 0.08
cards--home team
Previous season red 0.11
cards--away team
Notes: This table presents the selected characteristics.
Weekend games are the games that took place from Fridays
to Sundays. Midweek games are the games that took place
from Mondays to Thursdays.
APPENDIX B: PROPENSITY SCORE ESTIMATION
Table A2 contains the detailed estimation results of the propensity
(after variable selection by the double-LASSO as described in the main
text).
TABLE A2
Estimation of Propensity Score (Mean Marginal Effects)
Midweek
Midweek with without
Variables Attendance Attendance
Game characteristics
Attendance (in 10,000s) 0.003 **
Attendance as share of capacity 0.172
of stadium
Attendance as share of capacity -0.392
of stadium squared
Attendance as share of capacity -0.008
of stadium if higher than 0.99
Africa Cup of Nations months -0.026 * -0.017
August or September 0.225 *** 0.249 ***
Ln of stadium capacity -0.114 * 0.020
Season after World Cup or 0.064 *** 0.058 ***
European Championship
Two months before World Cup or 0.130 *** 0.109 ***
European Championship
Difference between teams' 0.000 0.000
values 2 months before
World Cup or European
Championship
Rounds 1-6 -0.071 *** -0.073 ***
Team characteristics
Difference in promotion status 0.013 0.011
Difference in HHI in the middle 4.068 *** 3.783 ***
of season
Difference in new coach in the -0.036 * -0.031
middle of season
Difference in teams' 0.000 0.002
standardized values
Difference in teams' -0.001 0.001
median values
Squared difference in teams' 0.001 *** 0.001 ***
values
Squared difference in teams' 0.000 0.000
values in the end of season
Ln of difference of teams' ratio -0.097 *** -0.091 ***
of top 3 to ranked
9-11 players' values
Ln of difference of teams' ratio 0.069 ** 0.074 **
of top 11 to ranked 12-22
players' values in the
beginning of season
Squared difference of teams' -0.001 -0.001
ratio of top 11 to ranked
12-22 players' values
Difference in standardized -0.010 -0.011
values of teams if values
are positive
Ratio between teams' values -0.031 -0.034
if higher than 5
Ratio between teams' values if 0.037 0.021
higher than 3 in the middle
of season
Difference between teams' values -0.009 -0.014
if ratio between values is
higher than 2 in the middle
of season
Difference between previous 0.015 ** 0.010
season's corners in the end
of current season
Ln of previous season's 0.014 -0.012
corners--home team
Difference in previous season's 0.001 0.001
fouls in the end of current
season
Previous season's red -0.074 -0.091
cards--home team
Difference in previous season's 0.003 0.001
shots in the beginning of
current season
Difference in previous season's 0.002 0.004
yellow cards
Midweek
Midweek versus
Variables versus Sundays Mondays
Game characteristics
Attendance (in 10,000s) 0.012 ** 0.017 **
Attendance as share of capacity -0.366 -0.570
of stadium
Attendance as share of capacity -0.705 -0.678
of stadium squared
Attendance as share of capacity 0.027 -0.078
of stadium if higher than 0.99
Africa Cup of Nations months -0.075 -0.110
August or September 0.303 *** 0.334 ***
Ln of stadium capacity -0.409 ** -0.703 ***
Season after World Cup or 0.165 *** 0.216 ***
European Championship
Two months before World Cup or 0.293 *** 0.331 ***
European Championship
Difference between teams' 0.000 -0.001
values 2 months before
World Cup or European
Championship
Rounds 1-6 -0.146 ** -0.127
Team characteristics
Difference in promotion status -0.015 0.062
Difference in HHI in the middle 12.959 *** 12.208 ***
of season
Difference in new coach in the -0.147 ** -0.047
middle of season
Difference in teams' -0.028 0.002
standardized values
Difference in teams' 0.002 -0.014
median values
Squared difference in teams' 0.000 *** 0.000
values
Squared difference in teams' 0.000 0.000
values in the end of season
Ln of difference of teams' ratio -0.265 *** -0.273 ***
of top 3 to ranked
9-11 players' values
Ln of difference of teams' ratio 0.136 0.324 ***
of top 11 to ranked 12-22
players' values in the
beginning of season
Squared difference of teams' -0.001 -0.010 **
ratio of top 11 to ranked
12-22 players' values
Difference in standardized -0.013 -0.049
values of teams if values
are positive
Ratio between teams' values -0.085 0.062
if higher than 5
Ratio between teams' values if -0.040 0.077
higher than 3 in the middle
of season
Difference between teams' values -0.004 0.080
if ratio between values is
higher than 2 in the middle
of season
Difference between previous 0.064 *** 0.022
season's corners in the end
of current season
Ln of previous season's -0.008 0.160
corners--home team
Difference in previous season's 0.016 -0.023
fouls in the end of current
season
Previous season's red -0.274 0.422
cards--home team
Difference in previous season's 0.012 -0.001
shots in the beginning of
current season
Difference in previous season's 0.009 0.030
yellow cards
Notes: Mean marginal effects presented. Inference based
on bootstrapping (99 replications) standard deviation and
using asymptotic normal distribution for inference. The
difference and ratio-related variables are always defined
as difference or ratio between home and away measures.
* Significant at 10%; ** significant at 5%;
*** significant at 1%.
APPENDIX C: RESULTS OF ROBUSTNESS CHECKS
Table A3 shows the results for using different comparisons between
midweek games and different definitions of weekend games. More
specifically, it shows the comparisons to all weekend games, games
played on Sunday, and games played on Friday. Furthermore, in comparison
to all weekend games the results are shown when controlling and not
controlling for attendance (which may be considered as affected by the
midweek effect and thus be endogenous).
TABLE A3
Effects of Different Specifications
Midweek Midweek
with without
Variables Attendance Attendance
Points--home team -0.286 *** -0.159
Points--away team 0.354 *** 0.207 **
Difference in points -0.640 *** -0.330 *
Goals--home team -0.204 ** -0.099
Goals--away team 0.047 0.039
Difference in goals -0.251 * -0.138
Shots--home team -0.719 ** -0.588 *
Shots--away team -0.295 -0.253
Difference in shots -0.424 -0.334
Shots on target--home team -0.502 *** -0.487 ***
Shots on target--away team 0.102 0.118
Difference in shots on target -0.604 ** -0.605 ***
Fouls--home team -1.260 *** -0.678 **
Fouls--away team -0.463 * -0.485 *
Difference in fouls -0.797 ** -0.194
Yellow cards--home team 0.158 ** 0.168 **
Yellow cards--away team -0.048 -0.160 *
Difference in yellow cards 0.206 0.328 **
Red cards--home team -0.021 * 0.010
Red cards--away team 0.039 0.020
Difference in red cards -0.060 * -0.010
Attendance as share of -0.083 ***
capacity of stadium
Remaining observations 83 96
in common support (%)
Midweek Midweek
versus versus
Variables Sunday Friday
Points--home team -0.319 *** -0.100
Points--away team 0.401 *** 0.192 *
Difference in points -0.720 *** -0.292
Goals--home team -0.224 *** 0.054
Goals--away team 0.128 0.288 ***
Difference in goals -0.352 *** -0.235 *
Shots--home team -1.049 *** -0.869 **
Shots--away team 0.076 0.296
Difference in shots -1.125 * -1.166 *
Shots on target--home team -0.725 *** -0.325 *
Shots on target--away team 0.175 0.545 ***
Difference in shots on target -0.900 *** -0.870 ***
Fouls--home team -0.908 *** -1.140 ***
Fouls--away team 0.078 -0.734 *
Difference in fouls -0.986 ** -0.407
Yellow cards--home team 0.213 *** 0.169 *
Yellow cards--away team -0.003 -0.115
Difference in yellow cards 0.216 * 0.284 **
Red cards--home team 0.011 0.033 *
Red cards--away team 0.026 0.030
Difference in red cards -0.015 0.004
Attendance as share of
capacity of stadium
Remaining observations 86 85
in common support (%)
Notes: Mean marginal effects presented. Inference based
on bootstrapping (4,999 replications for the first two
columns and 1,999 replications for the second two columns)
p values. The difference-related variables are defined as
difference between home and away measures. In all the
specifications, we control for attendance-related variables
except for column 2. In the first two columns, we compare
midweek matches to weekend matches by using all the data.
In column 3, we compare midweek matches to matches played
on Sunday. In the last column, we compare midweek matches
to matches played on Friday.
* Significant at 10%; ** significant at 5%;
*** significant at 1%.
APPENDIX D: LIST OF SOURCES
www.uefa.com
www.fifa.com
www.transfermarkt.com
www.football-data.co.uk
www.rsssf.com
www.espnfc.com
https://en.wikipedia.org/wiki/Bundesliga
www.regionalstatistik.de
REFERENCES
Belloni. A., V. Chernozhukov, and C. Hansen. "High-Dimensional
Methods and Inference on Structural and Treatment Effects." Journal
of Economic Perspectives, 28(2), 2014, 29-50.
Benabou, R.. and J. Tirole. "Self-Confidence and Personal
Motivation." Quarterly Journal of Economics, 117, 2002, 871-915.
Bernheim, B. D., and R. Thomadsen. "Memory and
Anticipation." The Economic Journal, 115(503), 2005, 271-304.
Bodory, H., L. Camponovo, M. Huber. and M. Lechner. "The
Finite Sample Performance of Inference Methods for Propensity Score
Matching and Weighting Estimators." Mimeo, University of St.
Gallen, 2016.
Bryson, A., B. Frick, and R. Simmons. "The Returns to Scarce
Talent Footedness and Player Remuneration in European Soccer."
Journal of Sports Economics, 14(6), 2013, 606-28.
Buhlmann. P.. and S. van de Geer. Statistics for High-Dimensional
Data: Methods, Theory and Applications. Heidelberg, Germany: Springer,
2011.
Buraimo, B. "Stadium Attendance and Television Audience Demand
in English League Football." Managerial and Decision Economics,
29(6), 2008, 513-23.
Buraimo, B., and R. Simmons. "Uncertainty of Outcome or Star
Quality? Television Audience Demand for English Premier League
Football." International Journal of the Economics of Business,
22(3), 2015, 449-69.
Coates, D., B. Frick, and T. Jewell. "Superstar Salaries and
Soccer Success: The Impact of Designated Players in Major League
Soccer." Journal of Sports Economics, 17(7), 2016, 716-35.
Cohen-Zada, D., A. Krumer, and Z. Shtudiner. "Psychological
Momentum and Gender." Journal of Economic Behavior &
Organization, 135, 2017, 66-81.
Courneya, K. S., and A. V. Carron. "The Home Advantage in
Sport Competitions: A Literature Review." Journal of Sport &
Exercise Psychology, 14(1), 1992, 13-27.
De Bruin, W. B. "Save the Last Dance for Me: Unwanted Serial
Position Effects in Jury Evaluations." Acta Psychologica, 118(3),
2005, 245-60.
Dohmen, T., and J. Sauermann. "Referee Bias." Journal of
Economic Surveys, 30(4), 2016, 679-95.
Forrest, D., and R. Simmons. "New Issues in Attendance Demand:
The Case of the English Football League." Journal of Sports
Economics, 7(3), 2006, 247-66.
Franck, E., and S. Nuesch. "Talent and/or Popularity: What
Does It Take to Be a Superstar?" Economic Inquiry, 50(1), 2012,
202-16.
Frick, B. Salary Determination and the Pay-Performance Relationship
in Professional Soccer: Evidence from Germany. Sports Economics After
Fifty Years: Essays in Honour of Simon Rottenberg. Oviedo, Spain:
Ediciones de la Universidad de Oviedo, 2006, 125-46.
Garicano, L., I. Palacios-Huerta, and C. Prendergast.
"Favoritism under Social Pressure." Review of Economics and
Statistics, 87(2), 2005, 208-16.
Georganas, S., M. Tonin, and M. Vlassopoulos. "Peer Pressure
and Productivity: The Role of Observing and Being Observed."
Journal of Economic Behavior & Organization, 117, 2015, 223-32.
Gibbons, M. R., and P. Hess. "Day of the Week Effects and
Asset Returns." Journal of Business, 54(6), 1981. 579-96.
Glejser, H., and B. Heyndels. "Efficiency and Inefficiency in
the Ranking in Competitions: The Case of the Queen Elisabeth Music
Contest." Journal of Cultural Economics, 25(2), 2001, 109-29.
Hastie, T., R. Tibshirani, and J. Friedman. The Elements of
Statistical Learning: Data Mining, Inference, and Prediction. 2nd ed.
New York: Springer, 2009 (10th printing with corrections, 2013).
Huber, M., M. Lechner, and C. Wunsch. "The Performance of
Estimators Based on the Propensity Score." Journal of Econometrics,
175(1), 2013, 1-21.
Huntingford, F. A., and A. K. Turner. Animal Conflict. London. UK:
Chapman and Hall, 1987.
Imbens, G. W., and J. M. Wooldridge. "Recent Developments in
the Econometrics of Program Evaluation." Journal of Economic
Literature, 47(1), 2009, 5-86.
Klumpp, T., and M. K. Polborn. "Primaries and the New
Hampshire Effect." Journal of Public Economics, 90(6), 2006,
1073-114.
Knowles, G., K. Sherony, and M. Haupert. "The Demand for Major
League Baseball: A Test of the Uncertainty of Outcome Hypothesis."
The American Economist, 36(2), 1992, 72-80.
Krumer, A. "Best-of-Two Contests with Psychological
Effects." Theory and Decision, 75(1), 2013, 85-100.
Krumer, A., R. Megidish, and A. Sela. "First-Mover Advantage
in Round-Robin Tournaments." Social Choice and Welfare, 48(3),
2017, 633-58.
Lechner, M., R. Miquel, and C. Wunsch. "Long-Run Effects of
Public Sector Sponsored Training in West Germany."
Journal of the European Economic Association, 9(4), 2011, 721-84.
Mazur, A., A. Booth, and J. M. Dabbs Jr. "Testosterone and
Chess Competition." Social Psychology Quarterly, 55(1), 1992,
70-77.
Moskowitz, T., and L. J. Wertheim. Scorecasting: The Hidden
Influences behind How Sports Are Played and Games Are Won. New York:
Three Rivers Press, 2011.
Neave, N., and S. Wolfson. "Testosterone, Territoriality, and
the 'Home Advantage'." Physiology & Behavior, 78(2),
2003, 269-75.
Oberhofen H., T. Philippovich, and H. Winner. "Distance
Matters in Away Games: Evidence from the German Football League."
Journal of Economic Psychology, 31(2), 2010, 200-11.
Page, L., and K. Page. "The Second Leg Home Advantage:
Evidence from European Football Cup Competitions." Journal of
Sports Sciences, 25(14), 2007, 1547-56.
--. "Last Shall Be First: A Field Study of Biases in
Sequential Performance Evaluation on the Idol Series." Journal of
Economic Behavior & Organization, 73(2), 2010, 186-98.
Palacios-Huerta, I. "Tournaments, Fairness and the
Prouhet-Thue-Morse Sequence." Economic Inquiry, 50(3), 2012,
848-49.
--"Psychological Pressure on the Field and Elsewhere," in
Beautiful Game Theory: How Soccer Can Help Economics, Chapter 5.
Princeton, NJ: Princeton University Press, 2014.
Paul, R. J. "Variations in NHL Attendance: The Impact of
Violence, Scoring, and Regional Rivalries." American Journal of
Economics and Sociology, 62(2), 2003, 345-64.
Pettersson-Lidbom, P., and M. Priks. "Behavior under Social
Pressure: Empty Italian Stadiums and Referee Bias." Economics
Letters, 108(2), 2010, 212-14.
Pollard, R. "Evidence of a Reduced Home Advantage When a Team
Moves to a New Stadium." Journal of Sports Sciences, 20(12), 2002,
969-73.
Schofield, J. A. "Performance and Attendance at Professional
Team Sports." Journal of Sport Behavior, 6(4), 1983, 196-206.
Scoppa, V. "Fatigue and Team Performance in Soccer: Evidence
from the FIFA World Cup and the UEFA European Championship."
Journal of Sports Economics, 16(5), 2015, 482-507.
Siganos, A., E. Vagenas-Nanos, and P. Verwijmeren.
"Facebook's Daily Sentiment and International Stock
Markets." Journal of Economic Behavior & Organization, 107,
2014, 730-43.
Sutter, M., and M. G. Kocher. "Favoritism of Agents--The Case
of Referees' Home Bias." Journal of Economic Psychology,
25(4), 2004, 461-69.
Terry, P. C., N. Walrond, and A. V. Carron. "The Influence of
Game Location on Athletes' Psychological States." Journal of
Science and Medicine in Sport, 1(1), 1998, 29-37.
Tibshirani, R. "Regression Shrinkage and Selection via the
Lasso." Journal of the Royal Statistical Society, Series B, 58(1),
1996, 267-88.
Van Ours, J. C., and M. A. van Tuijl. "In-Season Head-Coach
Dismissals and the Performance of Professional Football Teams."
Economic Inquiry, 54(1), 2016, 591-604.
Wang, C., D. Goossens, and M. Vandebroek. "The Impact of the
Soccer Schedule on TV Viewership and Stadium Attendance Evidence from
the Belgian Pro League." Journal of Sports Economics, January 17,
2016. doi: 10.1177/1527002515612875.
Ward, R. E. "Rituals, First Impressions, and the Opening Day
Home Advantage." Sociology of Sport Journal, 15, 1998, 279-93.
Wood, R. I., and S. J. Stanton. "Testosterone and Sport:
Current Perspectives." Hormones and Behavior, 61(1), 2012, 147-55.
Zou, H. "The Adaptive Lasso and Its Oracle Properties."
Journal of the American Statistical Association, 101(476), 2006,
1418-29.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online
version of this article:
Appendix S1. Results of additional robustness checks.
(1.) The Idol show is a popular reality television-music
competition format where judges and audience select the winner. For
additional information, see
https://en.wikipedia.org/wiki/Idols_(TV_series) (accessed April I,
2016).
(2.) On average, teams earn 1.6 points per game at home matches,
which is almost half a point more compared to playing away.
(3.) Note that the winning team receives three points, while the
losing team gets no points. In case of a draw, each team gets one point.
(4.) The elimination of the fatigue explanation is in line with
Scoppa (2015), who investigated all FIFA World Cups and UEFA
Championships and found no effect of additional rest days on the
teams' winning probabilities.
(5.) Territoriality is a protective response to an invasion of
one's perceived territory, which is common among animals. For
example, Huntingford and Turner (1987) showed that animals attack more
readily and with higher toughness when defending a home territory.
(6.) Numbers are taken from http://www
.bavarianfootballworks.com/2015/ll/23/9782628/how-much-will-bayern-munich-make-from-bundesliga-tvrevenues (accessed March 8, 2016).
(7.) See Krumer, Megidish, and Sela (2017) for additional details
on round-robin structure.
(8.) For example, in two other leagues with unbalanced allocation
of matches among different weekdays such as the English Premier League
and Championships division (second-highest division in England), in the
last nine seasons, 18.2% and 28.9% of the games took place on midweek
days, respectively. Moreover, after Saturday and Sunday, the third most
frequent day with regard to the number of games in the Premier League is
Wednesday. The corresponding day for Championships is Tuesday. Data are
available on www.football-data.co.uk (accessed December 5, 2016).
(9.) The relegation play-off format was introduced in the 2008-2009
season. Previously, up to the 2007-2008 season, three teams were
directly relegated to the Bundesliga 2.
(10.) More specifically, the data on the attendance are available
on www.transfermarkt.com, the data on within game measures, such as
shots, yellow cards, number of fouls, and so on, are available on
www.football-data.co.uk and www .espnfc.com. The data on regional
economic characteristics are available on www.regionalstatistik.de. And
the data on the coaches and stadiums were taken from
https://en.wikipedia.org/wiki/Bundesliga.
(11.) Assuming an equal probability of a loss, a win, and a draw.
(12.) See Coates, Frick, and Jewell (2016) for discussion on the
relationship between players' inequality in salaries and
teams' performance.
(13.) The results of the relevant regression analysis are available
upon the request.
(14.) See Van Ours and van Tuijl (2016) for discussion on the
effects of coach dismissals on team performance.
(15.) Note that the results would be symmetric if we focus on the
success of the away team instead.
(16.) For example, Georganas, Tonin, and Vlassopoulos (2015) found
some evidence that subjects being observed increase their productivity.
In addition, Bernheim and Thomadsen (2005) showed the importance of
behavioral implications of anticipatory emotions, whereas Benabou and
Tirole (2002) highlighted the role of memory in economic behavior.
(17.) For additional details on positive effect of testosterone on
performance in sports see the comprehensive review of Wood and Stanton
(2012).
ALEX KRUMER and MICHAEL LECHNER
Krumer: Post-doctoral Fellow, Swiss Institute for Empirical
Economic Research (SEW), University of St. Gallen, St. Gallen CH-9000,
Switzerland; Phone 0041-712-242-342, Fax 0041-712-242-302, E-mail
alexander.krumer@unisg.ch
Lechner: Professor, Swiss Institute for Empirical Economic Research
(SEW), University of St. Gallen, St. Gallen CH9000, Switzerland; CEPR,
London, UK; CESIfo, Munich, Germany; IAB, Nuremberg, Germany; IZA, Bonn,
Germany. Phone 0041-712-242-814, Fax 0041-712-242-302, E-mail
michael.lechner@unisg.ch
doi: 10.1111/ecin.12465
TABLE 1
Descriptive Statistics of Selected Variables
Weekend Games Midweek Games
Variable (N = 1,861) (AT = 152)
Game outcomes
Points home team 1.60(1.31) 1.36(1.37)
Points away team 1.14(1.26) 1.46(1.38)
Goals home team 1.56(1.26) 1.51 (1.30)
Goals away team 1.23(1.14) 1.39(1.21)
Shots home team 14.38 (4.97) 13.99 (5.20)
Shots away team 11.93 (4.59) 11.82 (4.52)
Shots on target home team 5.13(2.56) 5.01 (2.63)
Shots on target away team 4.21 (2.36) 4.41 (2.35)
Fouls home team 15.91 (4.52) 14.72 (4.34)
Fouls away team 17.02 (4.77) 16.34 (5.06)
Yellow cards home team 1.65(1.19) 1.70(1.14)
Yellow cards away team 1.95 (1.23) 1.86(1.31)
Red cards home team 0.07 (0.28) 0.09 (0.29)
Red cards away team 0.10(0.31) 0.10(0.32)
Game characteristics
Attendance 40,795 (15,882) 40,478(17,572)
Attendance as share of 0.91 (0.12) 0.86 (0.14)
capacity of stadium
Africa Cup of Nations 0.12 0.06
(dummy variable for
respective month)
Two months before World Cup 0.12 0.18
or European Championship
Two months after World Cup 0.07 0.24
or European Championship
Distance between the cities 366(182) 372(181)
of the teams (km)
Team characteristics
Value of home team 86.7 (65.4) 109.7(117.5)
(Mil. [euro])
Value of away team 90.3 (69.7) 103.9 (87.7)
(Mil. [euro])
Standardized value of -0.19(0.80) -0.02(1.03)
home team
Standardized value of -0.15 (0.80) 0.00 (0.99)
away team
Ratio of top 3 to ranked 2.34 (0.67) 2.18(0.51)
9-11 players' values--home
team
Ratio of top 3 to ranked 2.34 (0.65) 2.48 (0.71)
9-11 players' values--away
team
First season after 0.16 0.15
promotion--home team
(dummy variable)
First season after 0.15 0.11
promotion--away team
(dummy variable)
New coach--home team 0.18 0.11
(dummy variable)
New coach--away team 0.18 0.14
(dummy variable)
Notes: This table presents average values and standard
deviations (in brackets for nonbinary variables). The
difference and ratio-related variables are defined as
difference or ratio between home and away measures.
TABLE 2
Levels and Effects of Playing Midweek
Expected Value Expected Value
When Playing When Not Playing
Variables Midweek Midweek
Points home team 1.330 1.617
Points away team 1.489 1.135
Difference in points -0.159 0.482
Goals home team 1.384 1.588
Goals away team 1.284 1.237
Difference in goals 0.100 0.351
Shots home team 13.728 14.448
Shots away team 11.490 11.785
Difference in shots 2.238 2.662
Shots on target home team 4.696 5.198
Shots on target away team 4.322 4.220
Difference in shots on target 0.373 0.977
Fouls home team 14.534 15.795
Fouls away team 16.478 16.941
Difference in fouls -1.944 -1.147
Yellow cards home team 1.750 1.592
Yellow cards away team 1.880 1.928
Difference in yellow cards -0.131 -0.336
Red cards home team 0.047 0.068
Red cards away team 0.137 0.099
Difference in red cards -0.091 -0.030
Effect of Standard Error
Variables Playing Midweek of the Effect
Points home team -0.286 *** 0.098
Points away team 0.354 *** 0.095
Difference in points -0.640 *** 0.191
Goals home team -0.204 ** 0.095
Goals away team 0.047 0.087
Difference in goals -0.251 * 0.141
Shots home team -0.719 ** 0.345
Shots away team -0.295 0.287
Difference in shots -0.424 0.493
Shots on target home team -0.502 *** 0.166
Shots on target away team 0.102 0.123
Difference in shots on target -0.604 ** 0.200
Fouls home team -1.260 *** 0.295
Fouls away team -0.463 * 0.303
Difference in fouls -0.797 ** 0.370
Yellow cards home team 0.158 ** 0.070
Yellow cards away team -0.048 0.099
Difference in yellow cards 0.206 0.133
Red cards home team -0.021 * 0.020
Red cards away team 0.039 0.033
Difference in red cards -0.060 * 0.040
Notes: Average treatment effect. Inference based on
bootstrapping (4,999 replications) p values.
Difference-related variables are defined as difference
between home and away measures.
* Significant at 10%; ** significant at 5%;
*** significant at 1%.
COPYRIGHT 2018 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.