ec3144 undergraduate dissertation

52
An investigation of referee favoritism when allocating added time in English Premier League 2013 to 2015 Name: Rory O’Riordan Student Number: 113421072 Date: 03-05-2016 Module: EC3144 Supervisor: Dr. Robert Butler 1

Upload: rory-oriordan

Post on 20-Mar-2017

17 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: EC3144 Undergraduate Dissertation

An investigation of referee favoritism when allocating added time in English Premier League 2013 to 2015

Name: Rory O’Riordan

Student Number: 113421072

Date: 03-05-2016

Module: EC3144

Supervisor: Dr. Robert Butler

Research Question: Do referees behave favorably towards certain

principals in a football match in the English Premier League?

1

Page 2: EC3144 Undergraduate Dissertation

(I) Table of Contents Page

List of Figures 3

List of Tables 3

Abstract 4

Chapter 1: Introduction 5

Chapter 2: Literature Review 8

Chapter 3: Data Collection 14

Chapter 4: Methodology 16

4.1 Home Favouritism 17

4.2 Big Club Favouritism 19

Chapter 5: Results 21

Chapter 6: Discussion & Conclusions 30

References 34

2

Page 3: EC3144 Undergraduate Dissertation

(II) List of Figures

2.1-Extra timy by score margin (German Bundesliga 01/02)

(III) List of Tables

3.1- EPL 2013-2015 Descriptive Statistics

5.1 The Determinants of Additional Time in the EPL 2013-2015

5.2 The Determinants of Additional Time – Club Size 2013-2015

5.3 Determinants of Additional Time - Close Matches 2013-2015

5.4 Determinants of Additional Time - Close Matches & Club Size 2013-2015

3

Page 4: EC3144 Undergraduate Dissertation

(IV) ABSTRACT

This paper questions and examines the impartiality the decision making of referees regarding

FIFA’s Law 7- The Duration of the Match. This research includes all 760 games played in the

English Premier League over the course of two season; 2013/2014 and 2014/2015. We

investigate to see if home favouritism or a ‘big’ team bias exists when referees allocate

additional time at the end of a game. We found weak evidence that suggests referees display

favourable behaviour towards the home teams but we can confirm that there is a significant

bias towards ‘big’ clubs, suggesting that Fergie Time truly exits in the EPL. We furthered our

research by investigating close games (goal margin ≤1 at 90 minutes) and found no evidence

suggesting Fergie Time was present in these games. The results from this paper suggest that

while the concept of Fergie Time exists, its ability to change a match outcome is low.

4

Page 5: EC3144 Undergraduate Dissertation

1.INTRODUCTION

This dissertation will investigate referee decision making when allocating added

time/injury time at the end of games in the English Premier League (EPL). This paper

particularly focuses on whether EPL officials display a recurring bias in favour of the home

team and/or in favour of the ‘big’ clubs, defined by their financial and footballing

performance. It investigates the existence of this favouritism over the course of 760 EPL

matches form August 2013 until May 2015. There have been a number of empirical studies

carried out examining the existence of referee bias in top leagues around the world (Boyko, et

al., 2007, Buraimo, et al., 2010, Garciano, et al., 2005, Scoppa, 2008, Sutter & Kocher, 2004,

Pollard, 2008, Pollard, 2006, Pollard & Pollard, 1876-2003). This paper focuses on referee

behaviour strictly in the EPL. As well as focusing if referees displayed home favouritism,

this paper will investigate whether or not Fergie Time actually exists in the EPL.

The referees officiating matches in any league do not have total control over how

much added time is to be allocated. The Féderation Internationale de Football Association

(FIFA), the head authority in football, give guidelines to referees on how to calculate and,

therefore, grant the correct amount of time to be added at the end of each half. FIFA’s Law

7-The Duration of the Match is dedicated to give direction to match officials on how to award

the appropriate amount of time. The Law states that:

“An allowance is made in either period for the time lost through: substitutions,

assessment of players injuries, the removal of injured players form the field of play

for treatment, time-wasting, when the play is to stop for different reasons (e.g. critical

weather conditions, goalpost broken, floodlight failure. Many stoppages are natural

(e.g. throw-ins, goal kicks). An allowance is to be made only when these delays are

5

Page 6: EC3144 Undergraduate Dissertation

excessive. The referee shall not compensate for a timekeeping error during the first

half by increasing or reducing the length of the second half.

The announcement of the additional time does not indicate the exact amount of time

left in the match. Time may be increased if the referee considers it appropriate (i.e. if

there is time wasting during injury time) but never reduced” (FIFA, 2014, p.29).

The first line stated by FIFA on Law 7 states “The referee decides on the time lost in

each period” (FIFA, 2014, p.29). This clarifies that he allocates the amount of time his

discretion, not that of the linesmen, fourth officials or any other body officiating the game.

The media and previous research provide the reasoning for carrying out this

investigation on EPL referee behaviour. Refereeing decision making comes under constant

scrutiny by players, managers, pundits, journalists and basically, anyone with an interest in

football on a regular basis. They are often accused of giving decisions to the ‘big’ teams.

Many managers of the so-called lesser teams feel that the decisions seem to go against them

too regularly. This is where the coinage Fergie Time comes into context. Fergie Time is used

to describe the favouritism referees display towards ‘big’ teams when allocating added time.

The phrase is reference to former Manchester United Manager Sir Alex Ferguson who often

pressured and arguably intimidated match officials for greater amounts of added time. The

perception was that if his United teams weren’t winning, there would be enough time added

on to ensure they score a late decisive goal. This is a real life example of the principal-agent

problem, where the principal is the football team and the agent is the match official.

Referee’s display favourable behaviour towards one principal in a football match when there

are certain incentives in question.

6

Page 7: EC3144 Undergraduate Dissertation

There has been an abundance of research conducted the investigation of favouritism is

sport. It has repeatedly been discovered that favouritism in sport does truly exist but the

complexity of the situation still baffles researches. Pollard (1986) discovered favouritism has

been part of professional sport in England and North America since the 18 th century. Pollard

(2005) found that the magnitude of favouritism in association football was stronger in the

English Football League’s early years. But the reasons for the existence of favouritism in

sport is still an enigma to researchers in this area.

There has been a vast amount of research carried out investigating favouritism, but the

majority of the research has investigated home favouritism. There has been little research

investigating the presence of a bias towards the big teams. This paper classifies the status of

different principals by the teams financial and footballing performance which helps us

identify if agents display favourable behaviour towards certain principals. These officials are

under constant pressure and they are lambasted after every game. They are more often

criticised for their decisions rather than praised. These social pressures may play a part in the

referee’s decisions. This helps us get a better understanding to what effect a club’s reputation

has on the agent’s decision making, thus questioning the impartiality of referees in the EPL.

7

Page 8: EC3144 Undergraduate Dissertation

2.LITERATURE REVIEW

There have been studies carried out examining team advantages in the top leagues in

Europe: Serie A (Italy), Spanish Primera Liga, German Bundesliga and the English Premier

League. These leagues are comprised of the teams that annually contest for Europe,

footballs’s most prestigious club competition the Uefa Champions League. The teams

involved are identified as the strongest teams in their domestic competitions. They generally

have a larger financial backing and larger fan suppor. Recent literature has looked at home

advantade in terms of disciniplary decisions (Boyko, et al., 2007) (Buraimo, et al., 2010).

There is literature that focuses on officials being biased in their allocation of injury time

(Sutter and Kocher, 2004) (Garciano, et al., 2005) (Scoppa, 2008) (Rickman & Witt, 2008)

(Riedl, et al., 2015).

Boyko et al (2007) examined 5244 English Premier League games over the seaons

from 92/92 to 05/06 to test whether referees were swayed by crowd effects. They retrieved

teams involved, referee, score, attendence, yellow and red cards and penalty kicks converted.

The effect the crowd has on the referee is a common theme throughout these studies. They

found that referees were significanly affected by both the number of people in attendance and

crowd density as they peanalised the away team woth more yellow cards than the home team

and awarded the away teams more penalties. For every 10,000 person increase they found

home advanatge increased by approximately .086 goals. During this period they found a

negative relationship between refereee experience and home advanatge. With 50 referees

involved during this time, the refereees with greate expereince showcased less home

advantage.

Buraimo et al. (2010) examined matches in the Bundesliga and English top flight

8

Page 9: EC3144 Undergraduate Dissertation

from 2000 to 2006. They conducted a minute by minute bivariate probit analysis of bookes

and dismissals to detemine the probability of a caution at different times in a match. They

also found that away teams are awarded with more bookings which is indicative of home

team favouritisim as a reuslt of crowd pressure. During derby matches (mathes between

teams in the same area) they discovered that there was an increased probablility of cautions.

They also found that referees show a home team bias caused by crowd pressure:

“That the net effect of a running track is to increase cards issued to home players

suggests that the result is being driven by the referee's response to the proximity of

the crowd and this is consistent with referees typically being biased towards the home

team because of the presence of partisan spectators.”

They found Similar to Boyko et al.’s study, away teams received more yellow and red

cards than home teams. They provided rationale for these findings. They considered that the

away team are more often on the back foot defending and as a result, they are involved in

more tackles and that if the goal margin is larger , the number of bookings declines as

intensity evidently drops.

These two studies show how referees can be influenced by the crowd nois when

making decisions on sanctioning the players. The crowd noise and size is out of the control of

the referee and it has showed evidence to contribute to home advantage. An experiment was

undertaken where referees watched recorded natches without the sound on. Ther results

showed that referees called less fouls for the away team when crowd noise was on compared

to when it was just the video. (Nevill, Balmer, & Williams, 1999, 2002).

9

Page 10: EC3144 Undergraduate Dissertation

Sutter and Kocher (2004) analysed the Bundesliga during the 01/02 season. They

investigated the hypotheses related to injury time allocation: 1. Extra time in the second half

depends on the margin, 2. Extra time will be longer if the home team are trailing by 1 goal

than if it’s a draw or they are ahead by a goal and 3. Refereees add more time as the number

of spectators increase. They found evidence that supports all these hypotheses. This presents

referees expressing home team favoritism:

Fig. 2.1-Extra timy by score margin (German Bundesliga 01/02)

Source: Sutter and Kocher (2004)

They found that when the score margin is a single goal more time is played but when

the final outcome of the game was clear, less time is allocated. The crowd size and denisty

also contributed to referees being home team biased as more penalties were awarded to home

teams than away teams. An intereising discovery was that there was only 4 occasions when

goals scored in injury time altered the outcome of the match. The home team benfited from

these goals on 3 occasions while Bayern Munich (the Bundesliga’s most successful team)

were the only away team beneficiary.

Garciano et al. (2005) tested a similar hypotheses about referees favouring te home

team to satisfy the crowd. They examined how crowd effects referee behaviour in the Spanish

10

Page 11: EC3144 Undergraduate Dissertation

Primera Liga. They found similar resutls to Mattias and Kocher 2004: when the home team is

trailing by 1 goal, injury time is on average 35% above the average injury time added (3

minutes) but when the away team are ahead by a goal it is 29% below average. They also

found evidence that suggests referee bias is caused by crowd pressure. In games when the

attendance is larger the bias increases proving home favoritism as the home fan contingent is

usually larger. This was especially true in single goal margin games as the referees exhibted

this bias to a stronger magnitude.

Scoppa (2008) examined similar hypotheses to this dissertation in the Italian top tier, the

Serie A over the course of the two seaosns from 2003-2005. He investigated the existence of

home favouritism and a big club bias. He identified big teams by their economic, political

and media power. Scoppa examined injury time added on and also the poximity of the crowd

as a causal effect of referee favouritism when allocating additional time at the end of a game,

similar to Buraimo et al (2010). In the italian league abut 30 seconds extra was added on if

the home and/or big team were losing. Crowd proximity proved to be quite significant.

Crowd effects were stronger in stadiums where there was no running track separating the fans

and the pitch, thus the cue from the crowd shouting resulted in more fouls being called.

The studies done by Scoppa (2008), Mattias and Kocher (2004) and Garciano et al. (2005)

all found that crowd pressure plays a pivotol role in influecing referees, thus creating home

advantage. When the games are close coming towards the end the amount added on depends

on the current match result. When the home team were losing by one goal in all three leagues

more time was added on than if they were winning by a single goal, suggesting home

favouritism exists in the respective leagues. This gives the home team a greater chance of

11

Page 12: EC3144 Undergraduate Dissertation

improving their potential outcome and reduces the probability of the away team coming back

form a one goal deficit.

Studies carried out by Neil and Witt (2008) and Riedl et al. (2015) showed different

results in their studes. Neil and Witt (2008) examined Premier League and first

division.referees in 2001/2002 when referees were employed as professionals. A natural

experiment occurred showeing how financial incentives changed referees’ decion making.

There were two groups: the Select Group, 57 professional match officials who would receive

an annual retainer fee of £33000 and £900 per game, and the national list who weren’t

deemed professional. “The introduction of professional referees created financial rewards

for select groups of refs and this resulted in them allocating injury time more independently

than seen before in Garciano et al. 2005”. They found similar results to other studies

suggesting that when the score margin is larger at 90 minutes that less time is added on.

Riedl et al. (2015) are the most recent to have carried out this type of investigation.

They have looked at the German Bundesliga fixtures from 2000/01 to 2010/11. They

examined the ±1 goal margin at 90 minutes’ bias, whether time is added on so games end as a

draw rather than a team to win (charity bias) and they then examined do these two

hypotheses contribute to home advantage. They confirmed that ±1 goal difference bias does

exist but at a smaller scale (only 19 seconds (± 4) to be the difference) and that when leads

were more advantageous (by 2 or more goals at 90 minutes) less injury time was allowed.

They found evidence that showed favouring for the home team also through the charity bias.

20 seconds ±7 was added on when a potential goal in injury time would tie the game. In

terms of the home teams lead, as ΔG>0 is much more frequent than ΔG<0, this bias (charity

12

Page 13: EC3144 Undergraduate Dissertation

bias) favours the away teams. The effect of the biases was marginal and they were interpreted

to work in opposite directions in their favouring. They found no support that referee decision

on the length of injury time contributes to home advantage as the amount goals scored n

added time was small. This indicates there is no favouritism by referees in the Bundesliga

which contradicts previous studies conducted. They conducted a smaller time scale study on

the premier league from 2009-2013 and found that these two biases were present but the

effect was only marginal here too. The ±1 goal at 90 minutes bias caused a 13 second (±7)

difference in added time, while the charity bias caused on average a 16 second (±5) to injury

time. Only .03 additional goals for home teams were scored in injury time suggesting no

favouritism.

These studies by Riedl et al. (2015) and Neil and Witt (2008) show that referees may

neglect factors such as pressure from the crowd once financial incentives are involved. The

game of football has transformed as a whole. There is far more money involved in paying

players, managers, officials and far more revenue is generated for clubs meaning that there is

a greater loss/return from decisions going in/against a team’s favour. This suggests there is a

positive relationship between referees pay and their performance. Home advantage and

favouritism has reduced significantly in recent years according to Riedl et al. (2015)

suggesting the game has advanced and training for referees has improved.

13

Page 14: EC3144 Undergraduate Dissertation

3. DATA COLLECTION

In order to investigate the existence of referee biases in relation to the allocation of added

time in the English Premier League, there was data collected on every fixture during the

2013/2014 and 2014/2015 seasons. This dissertation is testing whether favouritism is

displayed towards two classifications of teams; home team favouritism and ‘big’ club

favouritism (‘Fergie Time’). In order to differentiate a ‘big’ club from the rest of the teams in

the league they must comply with a classification system. This paper defines a big club by

their financial and footballing performance. Thus, ‘big clubs’ must comply with the following

standards:

1. The club must be inside the top twenty worldwide clubs by revenue generation in the

Deloitte Football Money League Report for the two seasons being examined;

2013/2014 and 2014/2015.

2. The club must have participated in the Group Stages of the UEFA Champions League

and won a major domestic competition (the English Premier League, the FA cup

and/or the League Cup) in the past decade.

As the commercialization of football is ever increasing, it is important to judge a club on

their sporting exploits as well as their financial position. Any club which doesn’t meet the

criterion for a ‘big club’ will be known as a ‘small club’ hereafter. Only six EPL clubs met

14

Page 15: EC3144 Undergraduate Dissertation

the standards to be classified as a big club: Manchester United, Arsenal, Liverpool, Chelsea,

Manchester City and Tottenham Hotspur.

The dataset includes statistics from 760 EPL games which took place over the course of

two full seasons from August 2013 to May 2015. Data was collected for the matches using

the British Broadcasting Corporation (BBC) website. Fortunately, the data was obtained

before the BBC changed the format of their website. The changes they implemented resulted

in match reports not displaying how many seconds of additional time were played at the end

of the second half. Data was collected for each fixture on the teams involved, the amount of

added time allocated at the end of ninety minutes, the goal margin between the teams at the

end of ninety minutes of play, the total number of goals in each game, the total number of

yellow and red cards distributed in each match, the attendance, the referee officiating each

game and his age and experience and whether or not a serious injury occurred during the

game (a serious injury is said to have occurred if over six and a half minutes of added time

occurred). The other stoppages that occur throughout a game include the number of fouls,

corner kicks, throw ins and offside decisions. FIFA’s Law 7 states that these are natural

stoppages and that officials aren’t required to keep record of time elapsed during these

stoppages unless when the time elapsed is excessive. Table 2.1 displays descriptive statistics

for the two seasons in question.

Table 3.1 EPL 2013-2015 Descriptive StatisticsVariable Mean St. Dev. Min MaxAdditional Time (seconds) 262 80 6 1035Second half goals 1.33 1.18 0 6Margin after 90 minutes 1.36 1.16 0 6Substitutions 5.5 0.8 0 6Yellow Cards 3.52 2.00 0 10Red cards 0.17 0.47 0 6Referee Experience (Years) 7.64 4.34 0 15

15

Page 16: EC3144 Undergraduate Dissertation

Attendance 36,427 13,985 9100 75,454

16

Page 17: EC3144 Undergraduate Dissertation

4. METHODOLOGY

To investigate the presence and magnitude of favouritism in question in the 760 EPL

games in the sample, 14 regressions were calculated. Each regression was a simple linear

regression (OLS), corrected for heteroscedasticity. The dependent variable in each regression

was the amount of additional time in seconds. The independent variables include match

statistics mentioned earlier such as: number of second half goals, the goal margin at ninety

minutes, number of substitutions, yellow cards, red cards, the referee’s age and experience,

the log attendance and whether or not a serious injury occurred. The other dependent

variables were used to identify if referees behaved favourably towards the home teams and/or

big teams or if they were behaving adversely towards the away and/or small teams. The

match results in question refer to the outcome at the end of ninety minutes. It does not mean

the final result of the game as a decisive goal may have been scored during the injury time

added by the referee at the end of the second half.

Regressions (1) – (7) include all 760 EPL games from August 2013-May 2015.

Regressions (8) - (14) calculate the existence of favouritism in ‘close’ games. These games

are classified by the goal margin at ninety minutes. If the goal margin is 0 or 1 at the end of

normal time then it is classified as a close game, if the margin is greater than 1 than it isn’t

included. By comparing the magnitude of favourable behaviour in every game versus

favouritism in the close games we can test for the existence of some aspects of Fergie Time.

Regressions (1) – (3) and (8) – (10) are both testing for home advantage using the same

regression models. Regressions (4) – (7) and regressions (11) – (14) are both testing for ‘big’

club favouritism.

17

Page 18: EC3144 Undergraduate Dissertation

Sutter and Kocher (2004), Garciano et al. (2005) Scoppa (2008) and Riedl et al. (2015)

investigated the effect the goal margin has on the referee’s decision to allocate added time.

Riedl et al (2015) labelled this type of favouritism as a charity bias. They found similar

results which suggested there was a bias towards the home team in three of Europe’s top

league’s: German Bundesliga (Sutter and Kocher 2004, Riedl et al. 2015) , Italian Serie A

(Scoppa,2008) and Spanish La Liga (Garciano et al. 2005). They each found that when the

goal difference was greater than one at ninety minutes that less time would be added on as

opposed to when the margin is one or zero. Sutter and Kocher (2004), Garciano et al. (2005)

and Scoppa (2008) discovered more added time was allocated when the home team is behind

by one goal versus when they are ahead by one goal, thus providing evidence for Fergie

Time in their respective leagues.

4.1 Home Favouritism

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log( β7 A )+ β8 HW+β9 HL+β9+ε (1),(8)

Y t represents the additional time, in seconds, added by referees at the end of the second half

in each game. G is the amount of goals scored in the second half, M is the goal margin

between the two teams at ninety minutes, S is the number of substitutions made in the game,

yc is the number of cautions distributed by the referee in the game, rc represents the number

of red cards given in the match, I represents whether or not a serious injury occurred during

the second half, E is the referee’s experience officiating in the EPL in years and A represents

the attendance. The figure for attendance had to be given in a log form to erase problems with

heteroscedasticity. As the disparity between the stadium capacities in the EPL, it is better for

the OLS model to bring these values to scale rather than mix the high figures (e.g.

18

Page 19: EC3144 Undergraduate Dissertation

Manchester United vs. Chelsea, August 2013, Attendance: 75,032) with low figures (e.g.

QPR vs Hull, August 2014, Attendance:17603). The dependent variables mentioned already

in regression (1) are included in each regression. The dummy variable for regressions (1) -

(3), (8) – (10) is a ‘home draw’.

HW in regression (1) represents and home win and HL represents a home loss. These

dependent variables are used to identify whether the referees add different amounts of time

depending on the home team’s result at ninety minutes. The status of the club (big or small)

doesn’t matter here as we are only testing for home favouritism.

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log( β7 A )+ β8BHW +β9BHD+β9BHL+ε (2),

(9)

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log ( β7 A )+ β8 SHW +β9 SHD+β9 SHL+ε (3),

(10)

Regressions (2) and (3) include the club’s status classified by their financial and

footballing performance as mentioned earlier. In regression (2) BHW represents a big team

winning at home, BHD represents a big club drawing at home and BHL represents a big club

losing at home. There are only six clubs who qualify as a ‘big’ club. Regression (2) compares

their home matches to the rest of the games in the sample. In regression (3) SHW represents a

small club winning, SHD represents a home club drawing at home and SHL represents a

home club losing at home. Regression (3) is similar to regression (2) but considers the

opposite relationship i.e. compares small clubs home games versus the rest of the fixtures in

the sample. By comparing the amount of added time allotted when big teams are

winning/losing at home against when small teams are winning/losing at home it can help us

identify the existence of Fergie Time.

19

Page 20: EC3144 Undergraduate Dissertation

4.2 Big club favouritism

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log ( β7 A )+ β8 BBW +β9 BBL+ε (4)(11)

Regression (4) represents games when the six big teams (Manchester United,

Manchester City, Tottenham, Liverpool, Arsenal and Chelsea) play each other. The

independent variables here represent Big vs. Big win (BBW) and Big vs. Big loss (BBL). The

dummy variable for this regression is when the result is a draw at ninety minutes between two

big clubs

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log( β7 A )+ β8 BSW +β9 BSL+ε (5)(12)

Regression (5) considers when a big team played against a small team at home. BSW

considers a big club winning at home against a big team and BSL represents when a big club

is losing at home against a small team. The dummy variable foe regression (5) is when a big

club and small club are level at ninety minutes. This regression will should provide us with

more evidence on whether Fergie Time exists or not as the two principals involved represent

what Fergie Time refers to: a bias towards the big club.

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log ( β7 A )+ β8 SBW +β9 SBL+ε (6)(13)

Regression (6) examines the opposite to regression (5). For this regression the Small

team are at home against a big team; SBW representing a win for the home side at the end of

ninety minutes while SBL represents the small cub losing to a big team at ninety minutes. The

dummy variable for this regression is SBD, when the small club is drawing to a big side at

20

Page 21: EC3144 Undergraduate Dissertation

home. Similar to regression (5) , this will provide us with evidence supporting or negating the

Fergie Time hypotheses.

Y t=β0+β1G+β2 M +β3 S+β3 yc+β4 rc+β5 I+β6 E+ log( β7 A )+ β8SSW +β9 SSL+ε (7)(14)

The final regression testing for big club favouritism measures games involving only

small sides. The dummy variable in this case is the time in seconds added on when the two

sides are level at ninety minutes. Similar to this paper, Scoppa (2008) investigated for a big

team bias in Serie A. He identified big teams by their economic, political and media power

off the field in relation to the match fixing scandal. Serie A referee’s were favourable towards

the big teams in the Serie A when allocating added time. When the suspected teams were

losing, the referee’s added more time, which questions how impartial the Italian league

officials actually are. This gives more evidence that concept of Fergie Time exists not only in

the EPL but in other top League’s in Europe.

21

Page 22: EC3144 Undergraduate Dissertation

5. RESULTS

The F test (P>F-Value) for regressions (1) – (14) is significant to the 1% level. The F test was

0.000 for regressions (1) – (14). Table 5.1 shows the OLS results for the 780 EPL over the

two seasons. The R² value for regressions (1) – (3) suggests the model explains 44%-45% of

the variance in the amount of seconds added on by referees. As we can see many of the

independent variables are significant in explaining the reasons for the amount of added time

allocated at the end of the second half. The number of second half goals, the goal margin at

full time, yellow cards and serious injury all contribute to the amount of added time awarded

across the three regressions. As we can see, the goal margin is statistically significant in

negatively impacting the amount of time added on. This suggests that the greater the margin

is at the end of the second half, the referee reduces the amount of time added. Regression (1)

provides the first test for home favouritism. There is a greater amount of time added on

whether a home team is winning or losing at the end of the second half. Regression (1) found

that there is 34 seconds more added on when a home side is winning and 29 seconds extra

added on when they are winning. This provides evidence are impartial between home and

away teams as there is significantly more time added on whether a home team is winning or

losing.

Regression (2) considers matches when the big clubs are playing at home only and

compares them to the other matches in the sample. The results here are interesting. A

22

Page 23: EC3144 Undergraduate Dissertation

significant result was found that when a big club is winning (-11.46 seconds) or drawing (-

28.67 seconds) at home, that less time is allocated. Regression (3) examines the opposite

relationship to regression (2). A significant result found that when a small side is winning or

losing at home that more time is added on. This provides evidence that suggests referees are

impartial in their allocation of added time when small teams are at home.

If we compare these results from regression (2) and (3), there is evidence of Fergie Time

found in both set of results. The amount of time added on when a big team is winning at

home is significantly less than when a small team is winning at home.

5.1 The Determinants of Additional Time in the EPL 2013-2015Regression (1) (2) (3)Constant 283.02*** 223.46** 227.58**

(63.45) (79.64) (82.57)Goals 7.31*** 7.32*** 7.43***

(2.16) (2.17) (2.16)Margin -25.97*** -19.85*** -23.81***

(2.75) (2.17) (2.41)Substitutions 6.29** 7.51** 6.40**

(2.77) (2.82) (1.06)Yellow Cards 6.46*** 6.53*** 6.49***

(1.06) (1.07) (1.06)Red Cards 3.39 3.17 3.67

(4.79) (4.53) (4.67)Serious Injury 183.29*** 182.68*** 216.71***

(12.55) (19.77) (12.51)Referee Experience 0.59 0.53 0.64

(0.59) (0.59) (0.58)Log Attendance -19.97 -4.21 -4.66

(13.98) (17.72) (17.70)Home Win 28.68***

(7.96)Draw -

Home Loss 34.02***(8.44)

Big Club Home Winning -11.46*(6.4)

Big Club Home Drawing -28.67**(10.90)

Big Club Home Losing 14.57(9.23)

Small Club Home Winning 16.10*(4.57)

Small Club Home Drawing -14.08

23

Page 24: EC3144 Undergraduate Dissertation

(8.91)Small Club Home Losing 14.82*

(6.80)N 759 759 759Prob > F 0.000 0.000 0.000R² 0.4485 0.4403 0.4462VIF 1.45 1.26 1.47Statistically significant: ***at 0.1% level; **at 1% level; *at 5% level.† Results include referee fixed effects.†† The logarithm of the dependent variable (second half additional time in seconds) produces results that do not differ statistically from those presented and demonstrate robustness in the dependent variable.

In table 5.2 we see the results from regressions (4) – (7). These regressions

investigate the existence of a bias towards one of the principals in football matches for all

games over the two seasons, based on their status. Regression (4) considers the matches when

the six big clubs play each other only. The R2 value for this regression is strong at 70.32%.

AS we can see, five independent variables are statistically significant in explaining a change

in the additional time added on: the number of second half goals, the occurrence of a serious

injury, big home team winning or losing all contribute positively to the additional time,

whereas seen in the previous set of regressions, the goal margin negatively effects the amount

of time allotted. Regression (4) provides evidence which suggests referees are impartial in

their allocation of added time when two big teams are playing. There is a case which argues

that the referee is slightly more favourable to the big team playing at home because there is

15 seconds more time added on when the home side is losing against another big team

compared to when they are winning.

Regression (5) investigates an aspect of Fergie Time. Regression (5) solely deals with

games when a big club is at home to a small team. This subset amounts to 160 games over the

course of two seasons. The R2 value is 57.93%. Many of the recurring independent variables

are statistically significant in contributing to the increasing/decreasing the amount of seconds

added on: second half goals, the margin, the number of yellow cards and a serious injury. The

most interesting significant independent variable is the value for when a big team is losing at 24

Page 25: EC3144 Undergraduate Dissertation

home to a small team (p<0.1) which presents us with evidence which suggests the existence

of Fergie Time. When a big team is trailing a small team at home, an extra 30 seconds is

awarded. Big clubs do not play significantly more time when they are ahead or level at the

end of the second half. This finding suggests the referees are influenced by the characteristics

of the principals in a football match. As we can see from the results, the suggestion that

crowd effects impact referee decisions can be refuted. By profession, referees are meant to be

totally impartial between teams in a game but this paper suggests otherwise. There is no

reason big teams should be experiencing exclusive advantages.

Regression (6) looks at games where a small team is at home versus a big team. This

examines the opposite to regression (5). Only 38.12% of the variance in added time is

explained by regression (6). The same recurring independent variables as regression (5) are

statistically significant. Regression (6) actually provides evidence that referees are impartial

in their allocation of injury time during these games. The difference in time added on when a

small side is winning at home and when a small side is losing at home against a big team is

only 1 second. One conclusion can be drawn from the model is that when a small team plays

a big team at home that an extra half a minute will be played if either side are ahead.

Regression (7) is the final regression where all games over the two seasons are

included. As in regression (5) and (6) the same recurring independent variables are

statistically significant with the omission of second half goals. 46% of the OLS models

explains variance in the amount of time added on. Regression (7) examines games only

involving small clubs and it has the largest number of observations. Similar to regression (6)

25

Page 26: EC3144 Undergraduate Dissertation

the referees are more or less completely impartial. Significantly more added time (30

seconds) will be played whether the home team is losing or winning.

26

Page 27: EC3144 Undergraduate Dissertation

5.2 The Determinants of Additional Time – Club Size 2013-2015Regression (4) (5) (6) (7)Constant 193.15 50.40 179.08 286.02**

(336.22) (150.74) (183.242) (127.73)Goals 18.44** 7.49* 13.84** 0.42

(8.16) (3.98) (4.29) (3.65)Margin -42.76*** -24.67*** -31.52*** -21.14***

(6.44) (4.44) (5.42) (4.86)Substitutions 9.56 10.29 -0.04 9.47**

(8.02) (6.72) (5.49) (3.45)Yellow Cards 2.12 8.46*** 5.54** 6.90***

(4.02) (2.15) (2.50) (1.67)Red Cards 7.81 -7.50* 7.57 3.17

(10.58) (4.02) (11.21) (8.77)Serious Injury 268.48*** 193.24*** 119.98*** 210.72***

(15.93) (26.75) (15.54) (38.38)Referee Experience -0.79 0.51 -0.93 0.96

(1.57) (0.88) (1.08) (0.87)Log Attendance -5.11 24.61 13.74 -24.84

(68.10) (32.36) (41.00) (28.60)Big Vs. Big Win 87.21***

(23.24)Big Vs. Big Draw -

Big Vs. Big Loss 101.99***(21.69)

Big Vs. Small Win 12.69(15.72)

Big Vs. Small Draw -

Big Vs. Small Loss 30.25*(16.47)

Small Vs. Big Win 37.00*(22.41)

Small Vs. Big Draw -

Small Vs. Big Loss 35.50**(17.64)

Small Vs. Small Win 30.69**(10.71)

Small Vs. Small Draw -

Small Vs. Small Loss 30.38*(11.78)

N 60 160 175 363Prob > F 0.000 0.000 0.000 0.000R² 0.7032 0.5793 0.3812 0.46VIF 1.65 1.45 1.49 1.43

Statistically significant: ***at 0.1% level; **at 1% level; *at 5% level. † Results include referee fixed effects†† The log of the dependent variable produces results that do not differ statistically from those presented and demonstrate robustness in the dependent variable.

27

Page 28: EC3144 Undergraduate Dissertation

In all regressions, the experience of the referee and the number in attendance didn’t have a

statistically significant impact on the amount of added time. Regressions (2) and (5) can be

interpreted as evidence for Fergie Time. When we compare the results for regression (2) and

regression (3) we can see there is a bias in favour of the big teams when they are losing at

home as regression (3) negates the presence of home favouritism when the home team is a

small club. There isn’t enough proof to criticise referees for behaving favourable towards the

big clubs. Regressions (1), (4) and (6) and (7) actually provide evidence supporting EPL

officials’ impartiality. The time added on isn’t advantageous to either principal in question,

whether they are home/away and/or big/small. Regression (4) results can be argued that

referees behave favourably towards the home side.

Regressions (8) – (14) run the same tests but only on close games. The close game

factor (goal margin of ≤1) is something which may play a part on referees behaviour because

they are under more pressure. The margin factor is a key aspect of Fergie Time. The outcome

altering goals scored in additional time are quite low. Alex Ferguson often sought for more

time when his team could score a goal which would change the final outcome of a game in

his teams favour.

The independent variables substitutions, yellow cards and serious injury are

statistically significant in each regression (8) – (10). These set of regressions explain 38% -

39% of the added time allocated by referees at the end of the second half in close games.

Regression (8) suggests referees are favourable to the home team in close matches as 13

seconds extra time is played when they are behind. There is no statistically significant

evidence that suggests referees play more/less time is played when the home team is winning.

28

Page 29: EC3144 Undergraduate Dissertation

Regression (9) examines close matches when the six big clubs are at home. There is evidence

for Fergie time here because there is 17 seconds less played when they are winning at home.

Regression (10) suggests that when the small teams are playing, referees are impartial. In

these fixtures, there is significantly more time added on regardless of the outcome at ninety

minutes. If we compare the results for regression (9) and (10), we see that there is

significantly less time played when the big side is leading at home versus when the small

teams are leading at home at the end of the second half in close games.

Regressions (11) – (14) investigate the existence of Fergie Time in close matches

where the status of the principal is identified i.e. big or small. Regression (11) examines

games where the Manchester United, Arsenal, Chelsea, Liverpool, Manchester City, and

Tottenham play each other. This model is strong in explaining the causes of added time as the

R2 value is 68.60%. Regression (11) presents findings which show referees giving an

advantage to the home side in close games involving only big clubs. When the home team is

winning only 35 seconds extra will be played compared to when the home side is losing

where 83 seconds are played.

Regressions (12) examines the presence of Fergie Time when a big club is at home to

a small side. The model explains 52.61% of additional time awarded. There is no statistically

significant evidence that suggests referees behave favourably towards the big side in close

games. Regression (13) also does not find any evidence of a bias towards the big team or

home side when the small club is at home versus a big team when the margin is ≤1 at ninety

minutes. And finally, regression (14) does not suggest referees behave favourable towards

either side when there just small teams are involved

29

Page 30: EC3144 Undergraduate Dissertation

5.3 Determinants of Additional Time - Close Matches 2013-2015Regression (8) (9) (10)Constant 231.33* 142.63 146.18

(83.92) (103.69) (109.97)Goals 4.08 4.62 4.50

(2.92) (3.45) (2.89)Substitutions 10.63** 11.24*** 10.37**

(3.39) (3.45) (3.31)Yellow Cards 7.29*** 6.98*** 7.50***

(1.35) (1.36) (1.35)Red Cards -0.21 -0.80 -0.30

(6.86) (6.43) (6.60)Serious Injury 166.87*** 165.00*** 167.61***

(22.00) (21.94) (22.67)Log Attendance -12.78 8.24 5.43

(18.34) (22.94) (23.43)Home Win 6.54

(7.19)Draw -

Home Loss 13.45*(7.55)

Big Club Home Winning -17.10*(8.79)

Big Club Home Drawing -20.64*(11.31)

Big Club Home Losing 9.69(11.10)

Small Club Home Winning 17.45*(9.17)

Small Club Home Drawing 4.96(9.13)

Small Club Home Losing 15.13*(9.05)

N 473 475 470Prob > F 0.000 0.000 0.000R² 0.3863 0.3892 0.3863VIF 1.13 1.19 1.45

Statistically significant: ***at 0.1% level; **at 1% level; *at 5% level† Results include referee fixed effects.†† The log of the dependent variable produces results that do not differ statistically from those presented and demonstrate robustness in the dependent variable.

5.4 Determinants of Additional Time - Close Matches & Club Size 2013-201530

Page 31: EC3144 Undergraduate Dissertation

Regression (11) (12) (13) (14)

31

Page 32: EC3144 Undergraduate Dissertation

Constant -444.98 -103.57 245.11 167.79(609.49) (207.65) (230.04) (158.52)

Goals 10.24 5.37 11.34 -3.79(12.40) (5.01) (7.07) (5.05)

Substitutions 25.88* 14.52* 10.65 10.45*(13.23) (8.27) (9.14) (4.01)

Yellow Cards 2.26 11.92*** 4.98 7.61***(5.04) (3.12) (3.40) (2.03)

Red Cards 28.89 -9.42* 3.02 -3.20(21.37) (4.26) (16.61) (12.63)

Serious Injury 248.95*** 154.40*** 119.94*** 194.37***(26.67) (13.47) (17.78) (40.95)

Log Attendance 110.69 51.56 -14.66 2.63(120.37) (45.73) (54.00) (35.02)

Big Vs. Big Win 35.37*

(25.39)

Big Vs. Big Draw -

Big Vs. Big Loss 83.52***

(20.90)

Big Vs. Small Win -8.64

(15.88)

Big Vs. Small Draw -

Big Vs. Small Loss 3.51

(17.12)

Small Vs. Big Win 12.25

(24.18)

Small Vs. Big Draw -

Small Vs. Big Loss 4.28

(17.17)

Small Vs. Small Win 13.52

(8.96)

Small Vs. Small Draw -

Small Vs. Small Loss 15.90

(10.39)N 35 80 108 250Prob > F 0.000 0.000 0.000 0.000R² 0.6860 0.5261 0.2762 0.4443VIF 1.32 1.29 1.18 1.14

Statistically significant: ***at 0.1% level; **at 1% level; *at 5% level.† Results include referee fixed effects.†† The log of the dependent variable produces results that do not differ statistically from those presented and demonstrate robustness in the dependent variable.

6. DISCUSSION & CONCLUSIONS

32

Page 33: EC3144 Undergraduate Dissertation

This paper examines various factors which contribute to additional time in all EPL games

over the course of the 2013/2014 and 2014/2015 season. The impartiality of the EPL is

questioned through testing two hypotheses which suggest; (a) referees behave favourably

towards the home team and (b) referees behave favourably towards the big teams (Fergie

Time) when they decide on how much added time is appropriate to add on in the second half.

In the investigation, we can take away that certain recurring factors contribute to the

explanation of how much time is added on. These include the number of second half goals,

the goal margin between the teams at ninety minutes, the number of cautions and dismissals

awarded during a game, the number of substitutes made and second half serious injuries.

Evidence from the paper provides strong evidence supporting the Fergie Time

hypotheses, although there was weak evidence supporting the existence of home favouritism.

The results of regression (2) suggest that when a big club is winning at home, a significantly

less amount of time is played, this supports the existence of Fergie Time. The evidence

suggesting there is a home bias is relatively weak. Results from regressions (1), (3) and (10)

provide evidence which suggests that referees display no advantage to the home side.

This investigation provides evidence that there is a bias towards big clubs over small clubs in

relation to second half injury time. This concept is commonly referred to as Fergie Time in

the English media. It was discovered that big clubs play over a half a minute more when they

are losing home or away to smaller clubs.

33

Page 34: EC3144 Undergraduate Dissertation

Examining the impact that the goal margin has on referees’ decision making brought

about some interesting results. Regressions (8) – (14) considered games where the margin

was ≤1 at ninety minutes. Regressions (8) and (11) provide significant evidence which

suggests referees add more time when the home side are down by a goal. Regression (11)

considers games where only the six big teams are playing. It was discovered in this

regression, that 84 seconds more are played when the home team is down compared to when

they are level. There was no significant evidence which supported the existence of Fergie

Time in the close matches over the course of the two seasons. In games where the principals

were the same standard, regressions (4) and (7), referees were impartial when adding on time.

As referee experience and the number of people in attendance didn’t have an impact

on the amount of additional time played, we see different results to that found in the Serie A

(Scoppa, 2008). It was discovered that crowd noise and their proximity from the field of play

are the main cause of biased referee decisions. Similar to Scoppa’s (2008) paper though, we

see that there is evidence of favouritism towards the big teams. This paper found evidence

that supports Garciano et al (2005) paper on La Liga. This research found evidence that

suggests there may be a slight charity bias towards the home side when they are behind by

one goal in the EPL. Riedly et al. (2015) discovered this charity bias existed in the German

Bundesliga as well. He found that an extra 19 seconds is played when the margin is only a

single goal, whereas this paper found that the charity bias was towards the home team only in

close games.

There are some limitations to this paper. People with a keen interest in football may

be speculative of the six teams classified as big in this paper. There are arguments that other

34

Page 35: EC3144 Undergraduate Dissertation

teams included should be omitted and replaced by others. The method used to establish big

teams is appropriate in today’s football climate and can be replicated if investigating other

top tier leagues around the world to identify big teams. This paper only looks at two seasons

of the EPL which has been running since 1992. If it were possible to go back to the first full

EPL season in 1992/1993 and gather similar datasets, it would provide a greater amount of

evidence supporting or refuting the hypotheses questioned here. Future papers may include

international club competitions involving referees from England to test EPL referees

behaviour when teams from outside the United Kingdom are involved.

Solving the issue regarding added time is complex. There is no one right answer, but

if there were clearer directives to referees on how much they should allow to be added for

each stoppage, it would help make the game fairer and protect referees from criticism. If all

parties involved in football were provided with guidelines for how much added time should

be allotted for yellow cards, red cards, substitutions, goals etc. it would reduce the

uncertainty. It would be easier for the officials to appropriate added time and managers and

teams could then comprehend where the time is coming from. One solution, which is hasn’t

been mentioned is removing timekeeping duties from the referee completely. If there were a

third party, for example a television match official or a group of match officials away from

the field of play, put in charge of the allocation of added time. They would be away from the

field of play, therefore, they would be under less pressure from fans, players and managers.

The introduction of additional time at the end of each half has contributed to the excitement

and fairness in the game of football.

The officials are meant to be impartial and recent studies have proved evidence that the

FA may need to intervene to increase the transparency relating to how much time is the right

35

Page 36: EC3144 Undergraduate Dissertation

amount of time to allocate. If there were clearer directions given to match officials and if

referees followed them stringently, their impartiality could not be questioned.

36

Page 37: EC3144 Undergraduate Dissertation

References

Boyko, Ryan H., Adam R. Boyko and Mark Boyko (2007). Referee bias contributes to home advantage in English Premiership football. Journal of Sports Sciences, 25(11), 1185-1194.

Buraimo, Babatunde, David Forrest and Robert Simmons (2010). The 12th man?: refereeing bias in English and German soccer. Journal of the Royal Statistical Society: Series A (Statistics in Society), 173(2), 431-449.

Clarke, Stephen. R. and John M. Norman (1995). Home ground advantage of individual clubs in English soccer. The Statistician, 44, 509 –521.

FIFA (2014) The Laws of the Game, Fédération Internationale de Football Association, Zurich, Switzerland. [Retrived from http://www.fifa.com]

Garicano, Luis, Ignacio Palacios-Huerta and Canice Prendergast (2005). Favoritism under social pressure. Review of Economics and Statistics, 87(2), 208-216.

Nevill, Alan M., Sue M. Newell and Sally Gale (1996). Factors associated with home advantage in English and Scottish soccer matches. Journal of Sports Sciences, 14(2), 181-186.-Home Advantage

Nevill, Alan M., Nigel J. Balmer and Mark A. Williams (2002). The influence of crowd noise and experience upon refereeing decisions in football. Psychology of Sport and Exercise, 3(4), 261-272.

Nevill, Alan M., Tom Webb and Adam Watts (2013). Improved training of football referees and the decline in home advantage post-WW2. Psychology of Sport and Exercise, 14(2), 220-227.

Pollard, Richard (1986). Home advantage in soccer: A retrospective analysis. Journal of Sports Sciences, 4(3), 237-248.

Pollard, Richard (2006). Worldwide regional variations in home advantage in association football. Journal of Sports Sciences, 24(3), 231-240.

Pollard, Richard (2008). Home advantage in football: A current review of an unsolved puzzle. The Open Sports Sciences Journal, 1(1), 12-14.

Pollard, Richard and G. Pollard (2005). Long-term trends in home advantage in professional team sports in North America and England (1876–2003). Journal of Sports Sciences, 23(4), 337-350.

Reilly, Barry and Robert Witt (2013). Red cards, referee home bias and social pressure: evidence from English Premiership Soccer. Applied Economics Letters, 20(7), 710-714.

Riedl, Dennis, Bernd Strauss, Andreas Heuer and Oliver Rubner (2015). Finale furioso: referee-biased injury times and their effects on home advantage in football. Journal of Sports Sciences, 33(4), 327-336.

Rickman, Neil and Robert Witt (2008). Favouritism and financial incentives: a natural experiment. Economica, 75(298), 296-309.

Scoppa, Vincenzo (2008). Are subjective evaluations biased by social factors or connections? An econometric analysis of soccer referee decisions. Empirical Economics, 35(1), 123-140.

Sutter, Mattias and Kocher, Martin G. (2004). Favoritism of agents–The case of referees' home bias. Journal of Economic Psychology. 25(4): 461-469.

37