Experimental Effects of an Absent Crowd on Performances and Refereeing Decisions during COVID-19

Bryson, Alex; Dolton, Peter; Reade, J. James; Schreyer, Dominik; Singleton, Carl

Working Paper

Experimental Effects of an Absent Crowd on

Performances and Refereeing Decisions during

COVID-19

IZA Discussion Papers, No. 13578

Provided in Cooperation with:

IZA – Institute of Labor Economics

Suggested Citation: Bryson, Alex; Dolton, Peter; Reade, J. James; Schreyer, Dominik; Singleton, Carl

(2020) : Experimental Effects of an Absent Crowd on Performances and Refereeing Decisions during

COVID-19, IZA Discussion Papers, No. 13578, Institute of Labor Economics (IZA), Bonn

This Version is available at:

https://hdl.handle.net/10419/224020

Standard-Nutzungsbedingungen:

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen

Zwecken und zum Privatgebrauch gespeichert und kopiert werden.

Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle

Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich

machen, vertreiben oder anderweitig nutzen.

Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen

(insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten,

gelten abweichend von diesen Nutzungsbedingungen die in der dort

genannten Lizenz gewährten Nutzungsrechte.

Documents in EconStor may be saved and copied for your personal

and scholarly purposes.

You are not to copy documents for public or commercial purposes, to

exhibit the documents publicly, to make them publicly available on the

internet, or to distribute or otherwise use the documents in public.

If the documents have been made available under an Open Content

Licence (especially Creative Commons Licences), you may exercise

further usage rights as specified in the indicated licence.

DISCUSSION PAPER SERIES

IZA DP No. 13578

Alex Bryson

Peter Dolton

J. James Reade

Dominik Schreyer

Carl Singleton

Experimental Effects of an Absent

Crowd on Performances and Refereeing

Decisions during COVID-19

AUGUST 2020

Any opinions expressed in this paper are those of the author(s) and not those of IZA. Research published in this series may

include views on policy, but IZA takes no institutional policy positions. The IZA research network is committed to the IZA

Guiding Principles of Research Integrity.

The IZA Institute of Labor Economics is an independent economic research institute that conducts research in labor economics

and offers evidence-based policy advice on labor market issues. Supported by the Deutsche Post Foundation, IZA runs the

world’s largest network of economists, whose research aims to provide answers to the global labor market challenges of our

time. Our key objective is to build bridges between academic research, policymakers and society.

IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion. Citation of such a paper

should account for its provisional character. A revised version may be available directly from the author.

Schaumburg-Lippe-Straße 5–9

53113 Bonn, Germany

Phone: +49-228-3894-0

Email: [email protected] www.iza.org

IZA – Institute of Labor Economics

DISCUSSION PAPER SERIES

ISSN: 2365-9793

IZA DP No. 13578

Experimental Effects of an Absent

Crowd on Performances and Refereeing

Decisions during COVID-19

AUGUST 2020

Alex Bryson

UCL and IZA

Peter Dolton

University of Sussex and IZA

J. James Reade

University of Reading

Dominik Schreyer

WHU – Otto Beisheim School of Management

Carl Singleton

University of Reading

ABSTRACT

IZA DP No. 13578 AUGUST 2020

Experimental Effects of an Absent

Crowd on Performances and Refereeing

Decisions during COVID-19

The Covid-19 pandemic has induced worldwide natural experiments on the effects of

crowds. We exploit one of these experiments currently taking place over several countries

in almost identical settings: professional football matches played behind closed doors.

We find large and statistically significant effects on the number of yellow cards issued by

referees. Without a crowd, fewer cards were awarded to the away teams, reducing home

advantage. These results have implications for the influence of social pressure and crowds

on the neutrality of refereeing decisions.

JEL Classification: C90, D91, L83, Z20

Keywords: attendance, Coronavirus, COVID-19, home advantage,

natural experiments, referee bias, social pressure

Corresponding author:

Carl Singleton

Department of Economics

University of Reading

Whiteknights Campus, RG6 6UA

United Kingdom

E-mail: [email protected]

1. Introduction

It is uncommon to encounter truly experimental data outside of the laboratory, especially for

high-stakes outcomes. But the Covid-19 pandemic has created exceptional circumstances,

which allow us to contribute to the small literature investigating the effects of a home crowd

on the outcomes of sporting fixtures and referee impartiality (e.g., Dohmen and Sauermann,

2016; Garicano et al., 2005; Sutter and Kocher, 2004).

We exploit the natural experiment

arising from the Covid-19 induced shutdown of football seasons and the later resumption

mostly behind closed doors. Using data from 6,481 football matches played before and after

the mid-season shutdown in 17 countries, including 1,498 played without spectators, we find

that the absence of crowds reduced home advantage persistently, with the gap between home

and away team punishments significantly narrowing.

Past studies have shown that playing behind closed doors, in one-off matches, reduces

aspects of football home advantage (Pettersson-Lidbom and Priks, 2010; Reade et al., 2020).

They also suggested that the lack of social pressure from the crowd affects the referee, with

fewer punishments for foul play for the team playing away from home. But it is unclear from

one-off matches whether the driver of reduced home advantage was the lack of social pressure

or unfamiliarity with playing and officiating without a crowd.

Answers to these questions are of direct interest to the multi-billion-dollar sports industry,

because they inform understanding of the role that officials play. More broadly, those running

sports have a responsibility to the fans and others who pay substantial sums, either on season

tickets or TV subscriptions, to see high quality contests that are competitive and refereed

neutrally. The betting and financial markets are also interested in any margins associated with

sporting outcomes and the nature of referees’ decisions. Recent articles on football played

without crowds in the Economist and Financial Times are testament to the widespread interest

in these matters beyond the sports pages.

This study also contributes rare evidence from a

natural experiment on whether individuals make different and potentially biased decisions in

Covid-19 led to most professional football being completely suspended and then resumed without crowds. In

the English Premier League, for example, the last match was played on 9

March, then the sport was locked down,

and the 2019/20 season did not resume until 17

June. This 13-week hiatus was mirrored in many other countries

across Europe and further afield.

See ‘Graphic detail: Covid-19 and football’, in The Economist, 25

July 2020 and ‘Net benefit: Home

advantage in play but football refs are fairer’, in the Financial Times, 18

July 2020.

situations where there is some form of salient group membership (see the summary by Charness

and Sutter, 2012).

2. Data

Our main dataset contains 6,481 matches played in twenty-three professional leagues and

seventeen countries in the 2019/20 season (Table 1). These matches involved 369 football

teams and were officiated by 472 referees, with one team playing in their home stadium and

another team visiting. The studied leagues all had at least ten matches played without

spectators in the 2019/20 season.

In total, 1,498 (23%) of the matches were played behind

closed doors. Almost all matches played from mid-May 2020 had zero crowd attendance, with

exceptions in Australia, Denmark, Hungary, Poland, Serbia and Slovenia, where some matches

were played with small restricted crowds. Over all leagues, the average crowd attendance was

approximately 13,500 before the shutdown and just 200 after (but zero in 11 of the 17

countries).

The first two rows of Table 1 indicate that the share of matches ending in a home win fell

from 43.8% before the shutdown to 41.2% after. Figure 1 shows that the mean differences

between teams in the numbers of goals scored and yellow cards received within matches

decreased and increased, respectively, in most countries after the shutdown, suggesting that

home advantage was reduced.

Although other changes to the leagues could have affected

outcomes, such as the length of the mid-season break in training and allowing more in-match

substitutions, these differed across countries; the only common change was the effective

removal of stadium crowds.

Table 2 shows the simple mean differences in seven match outcomes, comparing those

played with and without a crowd in 2019/20.

Matches played behind closed doors were three

percentage points less likely to end in a home win (p-value< 0.1). In these matches,

significantly fewer yellow cards were awarded to the away teams for foul play by the referees

We dropped matches played in Turkey and Nicaragua due to attendance data being missing for many games.

See Appendix Figure B1 for the distribution of matches with fans and behind closed doors throughout the

2019/20 season.

Referees issue yellow cards as punishments to players deemed to have engaged in foul play, time-wasting or

dissent. If a player is issued with two yellow cards, then they are sent off and cannot return to the field of play.

Particularly serious offenses result in direct red cards.

See Appendix Figure A2 for the full distributions for home and away goals, and home and away yellow cards.

compared with when crowds were present, leading the gap between the home and away team

yellow cards, normally negative, to increase by around a third of a card (p-value< 0.01).

3. Estimation and Results

The raw differences described above do not control for the scheduling of leagues before and

after shutdown. It is also unclear whether these differences can be accounted for by the general

variation in crowd size (e.g., Buraimo et al., 2010), or from some disproportionate effect of

there being no crowd at all. For six different outcome variables (home win vs not, goal

difference, total cards, home and away yellow cards, and the difference between them), we

estimate the following using ordinary least squares (OLS):





= 





+ 





+ ℎ



+ 



+ 



+ 



, (1)

where y denotes the match outcomes; BCD is a dummy indicator that takes the value of one if

a match is played behind closed doors, and zero otherwise; ATT measures crowd attendance in

tens of thousands; h

and a

are fixed effects, capturing the home and away teams; and r

is a

referee fixed effect - these address the general tendency of some teams or referees, for example,

to earn and award more yellow cards. Subscripts are for: home team i, away team j, referee k,

and match m. The country and league fixed heterogeneities are absorbed by the sets of fixed

effects.

The results from estimating Equation (1) are presented in Table 3. They suggest that we

can explain between 25-34% of the variance depending on the outcome examined. Accounting

for team and referee heterogeneity and clustering the standard errors, neither playing behind

closed doors nor the regular variation in the size of the crowd significantly affected the

likelihood of a home win, the goal difference or the total goals scored in matches (

, Table 3,

columns I-III). Significantly fewer yellow cards were awarded to the away team without any

crowd at all (p-value< 0.01; column V), contributing to the gap in yellows between the home

and away teams narrowing by around a third compared with there being a crowd (p-value<

0.01; column VI). The normal variation in crowd attendance did not significantly affect this

gap. These results suggest that the total absence of the generally home-team-supporting crowd

reduced the social pressure on referees to punish the away team more harshly, leading to fairer

decisions. It is less likely that the mechanism behind this is a change in the performances of

players, since the final scorelines of matches were not significantly different without fans. We

also tested the sensitivity of these results to adding regressors in Equation (1) for the cumulative

number of matches played by the teams and officiated by the referees behind closed doors.

These were insignificant for all outcomes, suggesting that there was no re-familiarisation to the

home stadium surroundings with fans absent. This supports the conclusion that the lack of

social pressure from the home crowd was the cause of different punishment patterns, compared

with when crowds were present. The results support the conclusions from Pettersson-Lidbom

and Priks, 2010 and Reade et al., 2020, who only looked at one-off matches behind closed

doors.

The results are approximately identical when we estimate the Poisson model equivalents of

Equation (1) (Appendix Table A1). The significant reduction in the punishment gap between

home and away teams when the crowd was absent is robust to weighting each home team or

each country equally in Equation (1) (Appendix Table A2 & A3). We also add matches to the

dataset from the past five seasons in each league. This allows us to check the robustness of the

model specification. In Appendix Table A4, we find that the main result is robust to country-

month fixed effects; i.e., the main results are not driven by changes in play and decision-making

as football seasons reach their conclusions, when the consequences of individual performances

and decisions are clearer (e.g., winning championships or relegation). In Appendix Table A5,

we control for the fixed characteristics of matchups between home and away teams across

seasons. The main results are robust to the possibility that some matchups, e.g., local derbies

such as A.C. Milan vs Inter Milan, have different characteristics and may have taken place

more or less often behind closed doors. Finally, in Appendix Table A6, we re-estimate

Equation (1) for only the countries where small crowds returned after the shutdown, adding the

term 

 ∗ 



, where  is an indicator variable for the post shutdown period.

We find that 

is generally insignificant, such that there was no difference in how the normal

variation in crowds affected outcomes after the shutdown. In this limited set of countries, with

normally smaller crowds, the disproportionate effect of playing behind closed doors on away

yellow cards is insignificant. It is possible that the absence of social pressure may have been

felt less by the referees in these countries, since they normally work in front of relatively small

and sparse crowds anyway.

Fischer and Haucap (2020) found some evidence that the behind closed doors effect did decrease over time in

German football only after COVID-19.

4. Conclusion and discussion

In this study, we have used the natural experiment of football matches played behind closed

doors to retrieve estimates of the extent to which a crowd impacts on final outcomes and referee

decisions. We find that the absence of a partisan home crowd has no effect on the final match

scoreline, but it does result in a reduction of one-third of a yellow card for away teams relative

to home teams. We suggest that these results are causally due to a complete lack of crowd

pressure and the influence that this normally has on a referee to make decisions which favour

the home team.

Our findings are important for economics, not just for sports fans. Relatively little

experimental evidence exists about how a partisan crowd may influence outcomes in a way

which unfairly benefits some competitors. This suggests that the location of events could be

important. It justifies why neutral venues are often sought for the finals of key competitions,

with equal allocations of seats for the supporters of the participating individuals or teams.

Our paper also contributes to what is known about the way in which referees make

decisions. We have found causal evidence suggesting that they can be unfairly biased in favour

of one side or another by the presence of external crowds. This has implications for the judging

and citing of any competitive event when it is anticipated that the crowd could be partisan, for

example, in the Olympics (Balmer et al., 2003) or even in reality TV contests (Collins et al.,

2019). More generally, any judged contest with adversaries and a crowd present needs to

examine the fairness of the justice which may be administered.

A further implication of our findings is that they call into question the neutrality of referees

or arbitrators in the presence of a crowd. This means that we should be cognisant of the

possible influence that crowds can have on arbitrated, judged or refereed decisions.

References

Balmer, N., Nevill, A., and A. Williams. 2003. “Modelling home advantage in the Summer Olympic

Games.” Journal of Sports Sciences, 21(6): 469-478.

Buraimo, B., D. Forrest, and R. 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.

Collins, A., McKenzie, J., and L Vaughan Williams. 2019. “When is a talent contest not a talent

contest? Sequential performance bias in expert evaluation.” Economics Letters, 177: 94-98.

Charness, G., and M. Sutter. 2012. “Groups Make Better Self-Interested Decisions.” Journal of

Economic Perspectives, 26(3): 157–76.

Dohmen, T., and J. Sauermann. 2016. “Referee Bias.” Journal of Economic Surveys, 30(4): 679–

695.

Fischer, K., and Haucap, J. 2020. “Does crowd support drive the home advantage in professional

soccer? Evidence from German ghost games during the COVID-19 pandemic.” DICE

Discussion Papers 344, University of Düsseldorf, Düsseldorf Institute for Competition

Economics (DICE).

Garicano, L., I. Palacios-Huerta, and C. Prendergast. 2005. “Favouritism Under Social Pressure.”

The Review of Economics and Statistics, 87(2): 208–216.

Pettersson-Lidbom, P., and M. Priks. 2010. “Behavior under social pressure: Empty Italian

stadiums and referee bias.” Economics Letters, 108(2): 212–214.

Reade, J. J., D. Schreyer, and C. Singleton. 2020. “Echoes: what happens when football is played

behind closed doors?” Economics Discussion Papers em-dp2020-14, Department of

Economics, Reading University.

Sutter, M., and M. Kocher. 2004. “Favouritism of agents - The case of referees’ home bias.”

Journal of Economic Psychology, 25(4):461–469.

Tables

TABLE 1: Sample descriptive statistics for professional football in the 2019/20 season

Before shutdown After shutdown

Home win % 43.8 41.2

Mean attendance (1,000s) 13.5 0.2

Number of…

Teams 370 370

Referees 452 403

Leagues 23 23

Countries 17 17

Matches behind closed doors 73 1,425

All matches 4,915 1,566

Notes: author calculations using data from worldfootballdata.net, accessed 3/8/2020. See Figure 1 for a list of

the domestic leagues represented by each country and Appendix Table A7 for descriptives and sample sizes by

country. Mean attendance calculations include matches played behind closed doors, i.e., zero values.

‘Shutdown’ refers to the period from approximately mid-March to mid-May where no professional football was

played in these countries (see Appendix Figure A1).

TABLE 2 Differences in sample means, matches played behind closed doors vs with fans,

2019/20 season

Mean difference

(Behind closed doors - with crowd)

Home win share -0.03

Goal diff. (Home - Away) -0.07

Total goals 0.08

Home yellows 0.07

Away yellows -0.29

***

Yellows diff. (Home - Away) 0.36

***

Total yellows (Home + Away) -0.22

***

Notes: ***, **, * indicate significance from zero, i.e., no difference (behind closed doors minus with fans), at

1%, 5% and 10% levels, respectively, two-sided unpaired t-tests. Yellows includes second yellow cards. See

Table 1 and Figure 1 for dataset description. See Appendix Table A8 for these statistics and others by country.

TABLE 3: Estimated effects of playing football behind closed doors on match outcomes

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.026

-0.064

0.036

0.099

-0.221

***

0.320

***

(0.02)

(0.07)

(0.05)

(0.06)

(0.07)

Attendance (10,000s) (

)

-0.006

0.002

-0.004

0.037

0.079

***

-0.042

(0.01)

(0.04)

(0.03)

Home team fixed effects

Yes

Away team fixed effects

Yes

Referee fixed effects

Yes

0.281

0.336

0.240

0.316

0.302

0.240

6,316

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to three-way clustering (home team, away team,

referee) are displayed in parentheses. OLS estimates of Equation 1. Yellows includes second yellow cards. See Table 1 and Figure 1 for dataset description.

Figures

FIGURE 1: Average match differences between home and away team outcomes within professional football leagues, 2019/20 season, before and

after shutdown.

A. Goal Difference

B. Yellow Card Difference

Notes: author calculations using data from worldfootballdata.net, accessed 3/8/2020. Averages of Home minus Away outcomes over all matches in sample periods. Dashed

line is  = . Bubbles are proportional in area to the number of matches in the dataset in each country after 1

April 2020, see also Table 1. Leagues represented: Australia,

A-League; Albania, Superliga; Austria, Bundesliga and Bundesliga 2; Costa Rica, Primera Divisíon; Denmark, Super-liga; England, Premier League and Championship;

Germany, Bundesliga, 2. Bundesliga and 3. Liga; Greece, Super League; Hungary, OTP Bank Liga; Italy, Serie A and Serie B; Poland, Ekstraklasa; Portugal, Primeira Liga;

Romania, Liga 1; Serbia, SuperLiga; Slovenia, PrvaLiga; Spain, La Liga and Segunda Divisíon; Ukraine, Premier League.

Experimental effects of crowd absence on performances

and refereeing decisions during Covid-19

Alex Bryson, Peter Dolton, J. James Reade, Dominik Schreyer, Carl Singleton

Online Appendix

A. Additional Tables

TABLE A1: Estimated effects of playing football behind closed doors on match outcomes

(Poisson regression)

Total goals

Home yellows

Away yellows

(III)

(IV)

(V)

Closed doors (

)

0.011

0.045

-0.087

***

(0.03)

(0.02)

(0.03)

Attendance (10,000s) (

)

-0.003

0.018

0.045

***

(0.01)

Home team fixed effects

Yes

Away team fixed effects

Yes

Referee fixed effects

Yes

Pseudo R

0.063

0.085

0.076

6,316

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard

errors robust to three-way clustering (home team, away team, referee) are displayed in parentheses. Poisson

regression estimates of Equation 1. Yellows includes second yellow cards. See Table 1 and Figure 1 for dataset

description.

a.bryson@ucl.ac.uk, UCL’s Social Research Institute, UCL, 20 Bedford Way, London, WC1H 0AL.

P.Dolton@sussex.ac.uk, Department of Economics, University of Sussex, Brighton, BN1 9SL.

[email protected] & [email protected] (corresponding author), Department of Economics,

University of Reading, Whiteknights Campus, RG6 6UA, UK. dominik.sch[email protected]du, Center for Sports

and Management (CSM), WHU – Otto Beisheim School of Management, Erkrather Str. 224a, 40233

Düsseldorf, Germany.

TABLE A2: Estimated effects of playing football behind closed doors on match outcomes: equal home team weighting

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.032

-0.075

0.039

0.091

-0.215

0.306

***

(0.02)

(0.08)

(0.07)

(0.05)

(0.06)

(0.07)

Attendance (10,000s) (

)

-0.008

-0.000

-0.003

0.032

0.082

***

-0.049

(0.01)

(0.04)

(0.03)

Home team fixed effects

Yes

Away team fixed effects

Yes

Referee fixed effects

Yes

0.293

0.349

0.251

0.324

0.308

0.250

6,316

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to three-way clustering (home team, away team,

referee) are displayed in parentheses. OLS estimates of Equation 1, with observations weighted according to √(







∑





), where 



= 1/



is the inverse of the total

number of matches in the sample played by the home team in their own stadium and 



is the number of distinct home teams; we weight teams equally in the regression.

Yellows includes second yellow cards. See Table 1 and Figure 1 for dataset description, and Table 3 for comparable estimates without weighting.

TABLE A3: Estimated effects of playing football behind closed doors on match outcomes: equal country weighting

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.054

-0.148

0.134

0.085

-0.158

0.242

***

(0.03)

(0.09)

(0.10)

(0.06)

(0.07)

(0.08)

Attendance (10,000s) (

)

-0.019

-0.043

0.012

0.061

0.150

***

-0.089

(0.01)

(0.04)

(0.03)

(0.04)

(0.05)

Home team fixed effects

Yes

Away team fixed effects

Yes

Referee fixed effects

Yes

0.329

0.377

0.254

0.334

0.320

0.272

6,316

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to three-way clustering (home team, away team,

referee) are displayed in parentheses. OLS estimates of Equation 1, with observations weighted according to √( 







∑





), where 



= 1/



is the inverse of the total

number of matches in the sample played in country c and 



is the number of distinct countries; we weight countries equally in the regression. Yellows includes second

yellow cards. See Table 1 and Figure 1 for dataset description, and Table 3 for comparable estimates without weighting.

TABLE A4: Estimated effects of playing football behind closed doors on match outcomes, 2015/16 to 2019/20 seasons: controlling for

seasonality

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.087

***

-0.167

-0.147

0.221

***

-0.110

0.330

***

(0.02)

(0.09)

(0.10)

(0.07)

(0.08)

Attendance (10,000s) (

)

-0.007

0.035

-0.022

0.091

***

0.087

***

0.004

(0.01)

(0.03)

(0.04)

(0.02)

(0.03)

(0.02)

Home team-season fixed effects

Yes

Away team-season fixed effects

Yes

Referee fixed effects

Yes

Country-month effects

Yes

0.257

0.326

0.215

0.311

0.299

0.230

25,369

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to three-way clustering (home team, away team,

referee) are displayed in parentheses. OLS estimates of Equation 1, with the addition of country-month fixed effects, and home team-season and away team-season fixed

effects. Due to collinearity with the 2020 closed doors period for some countries, months May-July are combined as one period for the ‘country-month’ fixed effects. Yellows

includes second yellow cards. See Table 1 and Figure 1 for dataset description.

TABLE A5: Estimated effects of playing football behind closed doors on match outcomes, 2015/16 to 2019/20 seasons: controlling for matchups

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.069

***

-0.126

-0.066

0.239

***

-0.094

0.333

***

(0.03)

(0.09)

(0.07)

(0.09)

Attendance (10,000s) (

)

0.013

0.088

***

0.008

0.046

0.065

***

-0.018

(0.01)

(0.03)

(0.02)

Matchup fixed effects

Yes

Referee fixed effects

Yes

Country-month effects

Yes

0.433

0.476

0.398

0.465

0.463

0.412

25,399

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to two-way clustering (matchup, referee) are

displayed in parentheses. OLS estimates of Equation 1, with the addition of country-month fixed effects and replacing home and away team fixed effects with matchup fixed

effects, i.e., the unique combination of a home and away team, e.g., FC Barcelona hosting Real Madrid C.F. Due to collinearity with the 2020 closed doors period for some

countries, months May-July are combined as one period for the ‘country-month’ fixed effects. Yellows includes second yellow cards. See Table 1 and Figure 1 for dataset

description.

TABLE A6: Estimated effects of playing football behind closed doors or with restricted crowd numbers on match outcomes (Australia,

Denmark, Hungary, Poland Serbia and Slovenia)

Home win share

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

(I)

(II)

(III)

(IV)

(V)

(VI)

Closed doors (

)

-0.177

***

-0.501

0.129

-0.215

-0.112

-0.103

(0.06)

(0.26)

(0.21)

(0.16)

(0.14)

(0.23)

Attendance (10,000s) (

)

-0.012

-0.087

0.003

-0.036

0.523

-0.559

***

(0.05)

(0.14)

(0.19)

(0.13)

(0.20)

(0.19)

Attendance × COVID

(10,000s) (

)

0.021

0.312

-0.570

0.429

0.349

0.081

(0.18)

(0.48)

(0.34)

(0.43)

(0.45)

(0.53)

Home team fixed effects

Yes

Away team fixed effects

Yes

Referee fixed effects

Yes

0.340

0.362

0.248

0.342

0.319

0.318

1,147

Notes: ***, ** indicate significance from one at 1%, 5% and 10% levels, respectively, two-sided tests. Standard errors robust to three-way clustering (home team, away team,

referee) are displayed in parentheses. OLS estimates of Equation 1, with an additional term 

 ∗ 



. Yellows includes second yellow cards. See Table 1 and

Figure 1 for dataset description, and Table 3 for comparable estimates for all countries.

TABLE A7: Sample descriptive statistics for professional football leagues in the 2019/20 season: by country

Home win %

Mean attendance (1,000s)

Number of…

Number of matches

Before

shutdown

After

shutdown

Before

shutdown

After

shutdown

Teams

Referees

Leagues

Total

Behind

closed doors

After

shutdown

Albania

46.8

46.3

1.5

0.0

178

Australia

47.2

40.0

8.9

0.9

120

Austria

35.2

37.5

3.6

0.0

420

136

Costa Rica

48.6

50.0

2.5

0.0

242

Denmark

49.1

31.6

6.2

2.2

205

England

43.4

42.0

26.8

0.0

932

200

Germany

40.8

39.2

22.3

0.0

992

274

273

Greece

48.3

30.0

6.4

0.0

212

Hungary

41.2

48.3

3.2

2.9

177

Italy

42.9

43.1

15.3

0.0

760

247

225

Poland

50.0

34.2

9.1

2.6

246

Portugal

39.8

44.4

11.2

0.0

306

Romania

45.8

46.4

3.5

0.0

252

Serbia

51.4

52.0

2.0

1.2

233

Slovenia

44.8

30.9

1.4

0.1

180

Spain

43.1

42.4

18.2

0.0

841

232

231

Ukraine

44.9

44.7

4.0

0.0

185

All leagues

43.8

41.2

13.5

0.2

370

472

6,481

1,498

1,566

Notes: author calculations using data from worldfootballdata.net, accessed 3/8/2020. See Figure 1 for a list of the domestic leagues represented by each country. Mean

attendance calculations include matches played behind closed doors, i.e., zero values. ‘Shutdown’ refers to the period from approximately mid-March to mid-May where no

professional football was played in these countries (see Appendix Figure A1).

TABLE A8: Differences in sample means, matches played behind closed doors vs with fans, 2019/20 season: by country

Home win

Goal diff.

(Home-Away)

Total goals

Home yellows

Away yellows

Yellows diff.

(Home-Away)

Home reds

Away reds

Albania

-0.00

-0.02

0.83

***

0.03

-0.51**

0.53

-0.02

-0.00

Australia

-0.10

-0.19

0.55

***

-0.08

0.05

-0.13

-0.06

-0.08

Austria

0.02

-0.12

0.17

-0.08

0.26

0.03

0.01

Costa Rica

-0.02

0.22

0.30

0.21

-0.06

0.26

-0.05

Denmark

-0.26

-0.63

-0.70

0.39

-0.28

0.67

-0.06

-0.10

England

-0.01

-0.04

0.00

-0.13

-0.50

***

0.37

***

-0.02

0.00

Germany

-0.01

-0.02

-0.08

0.25

***

-0.24

0.49

***

-0.02

-0.03

Greece

-0.12

-0.36

-0.09

-0.14

-0.19

0.04

0.07

-0.18

Hungary

-0.24

-0.64

-0.27

-0.29

-0.89

0.60

0.10

-0.14

Italy

-0.00

-0.03

0.20

-0.14

-0.46

***

0.32

0.00

-0.07

Poland

-0.29

-0.77

0.53

-0.32

-0.12

-0.19

0.00

0.02

Portugal

0.05

0.17

0.26

0.28

-0.54

***

0.82

***

0.11

-0.01

Romania

-0.01

-0.06

-0.01

-0.17

0.10

-0.27

-0.01

Serbia

-0.05

0.49

0.59

-0.46

-0.44

-0.02

-0.10

0.08

Slovenia

-0.14

-0.77

0.10

-0.07

-0.90

0.82

0.07

-0.10

Spain

-0.01

-0.13

0.02

-0.47

***

0.49

***

0.03

-0.03

Ukraine

0.02

0.18

0.76

***

-0.47

-0.25

0.23

-0.17

-0.07

All leagues

-0.03

-0.07

0.08

0.07

-0.29

***

0.36

***

0.01

-0.02

Notes: ***, **, * indicate significance from zero, i.e., no difference (behind closed doors minus with fans), at 1%, 5% and 10% levels, respectively, two-sided unpaired t-

tests. Yellows includes second yellow cards. Reds includes straight red cards and second yellow cards. See Table 1 and Figure 1 for dataset description.

B. Additional Figures

FIGURE B1: Number of matches in the analysis by day, with fans and behind closed doors,

January to 3

August 2020, 2019/20 season

Notes: author calculations using data from worldfootballdata.net, accessed 3/8/2020. See Table 1 and Figure 1 in

the main text for further dataset description.

FIGURE A2: Distributions of home and away goals and yellow cards, 2019/20 season, with fans vs behind closed doors

A. Home Goals

B. Away Goals

C. Home Yellow Cards

B. Away Yellow Cards

Notes: author calculations using data from worldfootballdata.net, accessed 3/8/2020. See Table 1 and Figure 1 in the main text for further dataset description.