Bad for Business? The Effect of Hooliganism on English Professional Soccer Clubs R. Todd Jewell Rob Simmons Stefan Szymanski Abstract S

Bad for Business?

The Effect of Hooliganism on

English Professional Soccer Clubs 

R. Todd Jewell

Rob Simmons

Stefan Szymanski 

Abstract

Soccer hooliganism, defined as episodes of crowd trouble inside and outside soccer stadiums on match days, is commonly perceived to have adverse effects on the sport. However, it is also possible that hooliganism could have positive side-effects for clubs if it helps them to win games. In this paper we attempt to quantify both positive and negative effects on clubs, in terms of league performance and revenue generation. We measure hooliganism by arrests for soccer-related offenses. We analyze two distinct periods in the history of hooliganism in the top four divisions of professional English soccer: an early period during which hooliganism was a fundamental social problem (the seasons from 1984 to 1994), and a more recent period in which hooliganism has been less prevalent (1999 to 2009). In the 1984 to 1994 period we find evidence of a positive effect of arrests on league performance and an adverse effect on soccer club revenues. Both of these effects disappear in the more recent 1999 to 2009 period. Our results support the hypothesis that ‘gentrification” of the sport in recent years has reduced the amount of hooliganism and thereby has had a positive influence on revenue generation, as well as eliminating any benefit of hooliganism to clubs in terms of league performance.

 

Introduction

      Disorder at soccer stadiums – hooliganism, violence in the crowd and other forms of crowd disturbances – is sometimes called the English disease. While crowd violence at sports arenas has a long history (see e.g. Dunning, Murphy and Williams (1988)) and has been documented across the world (see e.g. Dunning, 2000), it has been synonymous with English soccer since it first became an issue of national concern in the 1960s, and international concern by the 1980s, culminating in the five-year ban of English clubs from European soccer competition following the Heysel tragedy of 1986. There is a large literature on hooliganism in sociology, criminology, and anthropology, which examines at great length the roots of soccer violence and the motives of the actors. There is also a small literature relating hooliganism to match day attendances. However, as far as we know there has been no research on the impact of hooliganism at soccer matches on the sporting performance and finances of soccer clubs.

      The lack of research on the effects of crowd violence on the sporting and business side of the game is perhaps surprising, given that hooliganism has been blamed for declining attendance at English soccer from the 1960s to the mid-1980s.¹ Moreover, given the extensive economic literature on favoritism under social pressure (e.g. Garicano et al. (2005)) influencing the outcome of competitive sports, one might imagine that fan violence could be a very potent form of social pressure. One explanation for this gap may be the limited research by economists on crowd disturbances. A second might be the paucity of data relating to crowd violence. This paper uses one of the few data sources available to examine the relationship between crowd disorder and team success, both on and off the pitch. From the 1984/85 season the total number of arrests by the police at soccer stadiums in England has been published for the 92 professional league clubs. Data for the four seasons 1995/96-1998/99 are missing, but is otherwise available up to the present day. Luckily, this gap in the data occurs at a natural breakpoint. Following the Hillsborough disaster of 1989 and the Taylor Report of 1990 English soccer stadiums were radically overhauled. The creation of the Premier League in 1992 led to a dramatic increase in club revenues especially at the top of the game, providing significant resources for stadium investment. As a result clubs were able to invest in significantly more powerful systems for the identification of those engaged in disorderly behavior (most notably the adoption of CCTV systems), while changes in the law made it easier to exclude known perpetrators from stadiums.² These new methods were being adopted just before the break in our data starts, and were fully in place by the time the break ends.

      We address two questions. First, how were arrests related to league performance over the season? Second, how were club revenues affected by arrests? In general, we find that crowd violence, as measured by the number of soccer-related arrests, was a decreasing function of the on-field success of the team, while club revenues were a decreasing function of arrests.

      However, the evidence strongly suggests that these relationships were most important before 1995. We find that a poorer performance in the league is associated with a higher level of arrests in the period prior to the 1995/96 season, suggesting that in part unrest was caused by fan disappointment with the performance of their team. In the post-1999 period this relationship is insignificant. Second, we find that club revenues in the pre-1995 period were decreasing in the recorded number of arrests, suggesting that a reputation for hooliganism affected the following of a club; once again this relationship was absent in the post-1999 period.

      Care must be taken in making a direct link from arrests to hooliganism, since the number of arrests can also be influenced by policing methods. Nonetheless, it is highly likely that the two are correlated. Based on an assumed correlation, our data and results tell a very plausible story. In the earlier era, which was the high point of hooligan concerns, there was a vicious circle running from poor team performance, to crowd disorder, leading to reduced support for the club and falling revenues, likely to lead to yet further performance deterioration on the field. In the later period, when tighter controls were imposed on the fans and rising ticket prices led to a degree of “gentrification,” these feedback mechanisms appear to have disappeared. ³ 

Hooliganism and Crowd Control in English Soccer

      The study of hooliganism has been almost exclusively the preserve of sociologists. Reviews of the historical record show that violent acts committed by soccer fans date back to the 19^th century (e.g. Dunning, Murphy and Williams (1988)), even if it only became a recognized social problem in England from around the start of the 1960s. Harrington (1968) produced the first government report on the subject and identified several dimensions of hooliganism: Rowdyism; horseplay and threatening behavior; foul support (e.g. throwing missiles); soccermania (collective misbehavior by mobs of 50 or so youths on the streets); Soccer riots (fights between rival fans); Individual reactions (e.g. assaulting policemen); vandalism; and physical injuries caused by fists, razors, knives, etc.

      In this paper, we use data on arrests at soccer stadiums as an indicator of the extent of disturbances. A report on public disorder at sporting event published in 1978 identified four principal laws under which the police could intervene – the Public Order Act of 1936 (threatening or abusive behavior in a public place), the Police Act of 1964 (obstructing a constable in the execution of his duty), the Prevention of Crime Act 1953 (having an offensive weapon in a public place without reasonable excuse), and the Criminal Law Act of 1977 (which increased the penalties for assault and resisting arrest). Further legislation followed (i) the Heysel Stadium disaster in 1985 in which 39 Juventus fans died when a wall collapsed as they tried to escape from an assault led from the section of the stadium reserved for Liverpool fans, (ii) the Bradford City disaster in which 56 fans died in a fire which started in the main stand (not hooligan related), and (iii) the Hillsborough disaster of 1989 in which 96 fans were crushed to death (not hooligan related).

      Part of this legislation was aimed at improving conditions inside stadiums. But hooliganism was also the focus. Thus, the Sporting Events (control of alcohol) Act (1985) made it an offence to consume alcohol at soccer stadiums and on trains or buses going to stadiums. The Football Spectators Act (1989) made it possible for the courts to ban fans from a stadium and to require an individual to surrender his passport if suspected of being a hooligan likely to travel to a game being played overseas. The Football Offences Act (1991) made it illegal, inter alia, for fans to enter the playing area, to engage in racist chanting, or to throw missiles at a soccer match. These laws were further strengthened by the Football (Offences and Disorder) Act (1999) and the Football (Disorder) Act (2000).

      While soccer clubs and the police have focused on prevention, primarily through the segregation of fans, increased surveillance, and heavy policing, the number of arrests at soccer matches suggests that they have not been wholly successful. But a by-product of legislation has been increasingly accurate quantification of hooligan related activities, at least insofar as they are measured by arrests.

      Sociological studies have generally tried to place hooliganism in the context of broader social problems.⁴ One of the earliest approaches by Taylor (1971) viewed hooliganism as a response by working class fans to the appropriation of soccer clubs by owners and directors intent on commercializing the game. In this explicitly Marxist framework, violence was triggered as a response to alienation from control of the club which traditionally rested with the fans. Not surprisingly, this approach was widely rejected since (a) it was not clear that a golden age of fan ownership ever existed, and (b) fan violence seemed to be largely aimed at other fans, not the owners. Others such as Marsh et al (1978) tried to derive a more anthropological view, and argued that much of what happened at soccer matches, while appearing disorderly to outsiders, actually represented a form of ritual that was highly codified and in fact involved much less violence than commonly thought.

      By the 1980s the evidence made it impossible to believe that real violence was not involved, but the ritualistic aspect was taken up by the so-called Leicester School (e.g. Dunning et al, 1988) which argued that hooliganism was a form of “aggressive masculinity.” Elements of codification and ritual helped to establish identities and a certain kind of social order, including a means to “promotion” for those engaged in hooliganism. This model has also come in for significant criticism on the grounds that hooliganism is clearly not restricted to the working class, and because of the school’s insistence that hooliganism was an ever present phenomenon, while most researchers have been prepared to accept that there was a clear shift in the 1960s. More pragmatic accounts have tended to focus on issues such as the decline of authority, greater individual freedom, and, to some extent, adverse economic conditions.

      There can be little doubt that the poor quality of stadiums did not help to engender a positive attitude, and antagonism between the police and fans was a two-way street. The tragedy at Hillsborough in 1989, where 96 fans were crushed to death while the police stood by and did little or nothing, perfectly illustrates the breakdown in social relationships. While subsequent inquiries have demonstrated conclusively that neither hooliganism nor excessive consumption of alcohol (often associated with crowd violence) contributed in any way to the tragedy, the police at the time, at senior levels, tried to put the blame on rowdy fans.⁵ More generally, some scholars have argued that the extent of hooliganism can actually be caused by police tactics, with some police forces offering an aggressive approach to alleged hooliganism (Pearson, 1999). The effects of more aggressive policing on hooliganism are a priori ambiguous. Poutvaara and Priks (2009) present Swedish evidence to show that discriminatory targeting of hooligan groups by police backed up by police intelligence can reduce hooligan activity. On the other hand, indiscriminate police tactics such as use of teargas to disperse unruly fans or random prison sentences for convicted fans could actually raise hooligan activity by enhancing the idea that the relationship between fans and police is necessarily antagonistic.

      An interesting aspect of the Harrington Report was a survey of about 2,000 readers of The Sun, a tabloid newspaper, of whom 90% claimed that hooliganism was an increasing problem. Two thirds of respondents believed that “supporting losing team” was a contributory factor, making it the most commonly cited reason (ahead of “needle match or league match”, “foreign influences”, “overcrowding on terraces”, “poor ground facilities”, or “supporting winning team”). Subsequent research has almost completely neglected this explanation. The reasons for this neglect are not entirely clear. Since there will always be losers, perhaps people feel uncomfortable that the explanation amounts to something which cannot be avoided. Or perhaps anger at defeat is too simple an explanation, and researchers are more concerned with the deeper question as to why defeat should trigger a violent reaction in some but not others. Or maybe hooliganism as a negative response to a losing team was deemed to be an untestable proposition. Certainly arrests data were not published until the mid-1980s, and it was only by the mid-1990s, when the problem was already coming under control, that a large enough time series for econometric testing had been generated.⁶

      As far as the adverse effects on clubs are concerned, there also seems to have been little interest in identifying winners and losers. One recent study (Rookwood  and Pearson,2012) did point out, interestingly, that most fans do not see hooligans as an entirely negative phenomenon, and identified benefits that non-hooligans might derive from the presence of hooligans- distraction, protection, and reputation. No one associated with a club could ever openly claim that they had benefited from hooliganism, even though this view is quite consistent with the well-established literature on favoritism under social pressure (Garicano et al, 2005).⁷

      We see fan violence as potentially exerting a short-run and long-run effect. In the short term the threat of aggression might intimidate rival fans (who are thus less vocal in supporting their team), officials (susceptible to social pressure) and possibly players on rival teams (who are threatened with violence). These effects might all increase the probability of winning and thus the standing of the team, which is likely to generate increased support. In the longer term, however, if a team comes to be associated with hooliganism it may lose support since non-violent fans may fear for their security, and its costs will increase since the police will insist on a heavier presence, and in the English system each club is required to pay for the policing at their own stadium.

      So far as we know we are the first to analyze the relationship between hooliganism and the financial performance of soccer clubs. At the very least, we know that clubs have to pay for the policing of their own games, and so the greater the perceived hooligan problem the greater the cost. Using a database of financial accounts we are able to examine the effects in detail. 

Methodology and Data

      Beginning with a basic model of revenue determination, we assume that club revenue is an increasing function of on-pitch performance (P) and a decreasing function of fan violence (V). Club revenue will also be influenced by market potential (M). The basic revenue relationship is given in equation (1), where i and t index clubs and seasons respectively. As stated above, ∂R/∂P > 0 and ∂R/∂V < 0. The direction of the effects of market size on club revenues will vary depending on the empirical measures employed.

1. Revenue_it = R(P_it, V_it, M_it)

We assume that fan violence is influenced by a club’s on- and off-field performance.⁸ Specifically, fans are less likely to engage in violent behavior at any given soccer match if the team is performing better.⁹ Equation (2) summarizes this relationship:

2. Violence_it = V(P_it) , where ∂V/∂P > 0

      The financial accounts of English football clubs can be obtained from Companies House, a UK government agency.¹⁰ Revenue_it is measured as total revenues for each club i in each season t valued in real UK millions of pounds. Since both revenues and wage bills have risen at a much greater rate than general inflation, use of a consumer price index to deflate revenues is inappropriate. We construct a football price index based on wage inflation. We take a team wage bill series compiled from club balance sheets. The wage bills are separated by division and a division-varying wage index is compiled with base year at 2009. This enables the wage index to reflect the greater rate of increase in wages and revenues in the top division, now the Premier League, than in lower divisions. Specifically, if clubs are win-maximizers then the only way to assess the real value of revenues is through league rank. The cost of league rank varies with the wage costs of the players hired. But this wage cost is rising much faster in the top division, most recently through the high growth rate of broadcast rights values, in turn transmitted directly into team payrolls with an approximately unit elasticity. If we were to deflate lower division revenues by an average wage cost across all divisions it would appear as if lower division real incomes are lower than they really are. Hence, we deflate Revenue by our division wage cost index to generate revenue in real English soccer terms.¹¹

      This paper relies on arrests data sourced from two places and two distinct eras. First, we have data by season and by club on the total number of arrests in and around football stadia in England and Wales over the seasons 1984/85 to 1994/95.¹² These data were supplied by the Association of Chief Police Officers, collated by the Sir Norman Chester Centre for Football Research at the University of Leicester and published in the Digest of Football Statistics. Our second data set covers a later period, 1999/2000 to 2009/10 and comprises United Kingdom Home Office arrests data, comparable to the data from the Digest of Football Statistics. We were unable to locate arrests data between the 1995/96 and 1998/99 seasons as these were not officially published.

      Our data cover the 92 clubs participating in the top four divisions of English professional soccer. Until 1992 the clubs in these four divisions constituted the Football League. After 1992, the top division broke away to form the Premier League (to gain control of all broadcast money generated in this division, while retaining the promotion and relegation relationship with the lower divisions), and so the Football League constituted only the second, third and fourth tiers of competition. Notwithstanding the issues of financial control, the four divisions represent a single sporting structure in both our periods, which we will refer to generically as Professional English Football (PEF).

      Although we would prefer to have access to the missing data we would nevertheless argue that the two sample periods represent quite distinct periods in terms of the evolution of football hooliganism in England and Wales. As mentioned above, the Hillsborough disaster, and the subsequent Taylor Report (1990), is commonly viewed as a defining moment in modern English football. In the aftermath of the disaster all-seater stadia were mandated and surveillance of fans and security was significantly tightened. These changes also coincided with the advent of satellite broadcasting which dramatically bid up the value of broadcast rights, brought unprecedented sums of money into game and signaled a change in the nature of fandom, often described as “gentrification.” Soccer in England became fashionable, ticket prices shot up, and clubs invested in significant stadium improvements to meet the demand. The data for our early period ends just when these changes are taking hold, and our later period starts by the time these changes well fully established.

      Given that our data are annual, we normalize the total number of arrests per season by the aggregate number of fans attending in that season (i.e. arrests per 1000 fans). Attendance ranges from 30,000 to 40,000 per season in the top division and 2,000 to 3,000 in the fourth tier. Thus in general the greatest number of arrests occurs at clubs with the highest average attendance, but in our view a higher arrest rate per fan signifies a greater problem for the club.

      We expect that our data are subject to measurement error. Biases are possible in either direction. Arrests could overstate the extent of hooliganism to the extent that police indiscriminately arrest fans as part of an over-zealous approach. But arrests involve considerable police effort in terms of processing charges. Police then have an incentive to only arrest fans if they really feel these fans have committed offenses. Then hooliganism may be understated by the number of arrests. In the absence of any detailed evidence, we simply assume that these biases are offsetting.

      Reconsider equations (1) and (2) in the presence of unobservable variables that influence both club revenue and the amount of fan violence. For instance, assume there exists an unobservable measure of “passion for your club,” which positively influences both revenue and arrests. Specifically, teams with more “passionate” fans will tend to have higher revenues, ceteris paribus, but these teams will also tend to see more fan violence as “passion” spills over into hooliganism. The existence of such unobservable measures in systems of equations leads to an estimation bias, termed “unobserved heterogeneity bias” or “endogeneity bias.” In the case of unobserved heterogeneity in fan passion, an OLS estimation of equation (1) would result in a positive bias on the coefficient on V in the revenue equation. Since we hypothesize that the ∂R/∂V will be negative, endogeneity bias will lead the coefficient on V to be biased toward zero, i.e., smaller in absolute value than the unbiased value.¹³

      We utilize 2SLS estimation to control for the potential endogeneity of fan violence. Specifically, we estimate a first-stage regression of equation (2) using logged values of Arrests_it and create fitted values for fan violence (PredLogArrests_it), which are used as instruments for fan violence in a first-stage regression of equation (1). 2SLS requires at least one exclusion restriction to identify fan violence separately from club revenue. In the case of soccer-related arrests, we can use lagged values of arrests per 1,000 fans per game as an exclusion restriction, as past values of arrests are indicative of a tendency toward or against fan violence.¹⁴

      On-pitch performance is measured as the negative log of season-ending position for club i in season t and for each division n (Rank_itn) and, defined this way, is expected to be positively related to team revenues (Szymanski and Kuypers, 1999; Szymanski and Smith, 1997). Past performance may influence club revenues, especially if the club in question was relegated or promoted at the end of the previous season. The variables LagPromote_it and LagRelegate_it are indicator variables that equal one if a club was promoted or relegated, respectively, in the previous season. The coefficients on LagPromote_it and LagRelegate_it cannot be signed a priori.

      Market potential is measured with several variables. First, we include fixed effects dummies for club, division, and season. Second, we expect that clubs with all-seater stadiums will generate greater revenues; Allseater_it equals one if the club played in such a stadium in season t. Third, we include indicators of the number of derby games in each season for each club. Derby1_itn equals one if club i has one traditional rival in division n during season t. Derby2_itn if a club plays more than one traditional rival in league play in season t. The coefficients on the derby indicators are expected to be positive, since rivalry games generally lead to greater attendance demand and should lead to greater revenues.¹⁵ 

      In the next section, we present results of IV regressions of equation (1) of the following general form, where , , and  are vectors of fixed effects for club, season, and division respectively and _it is a white noise error term.

3. Revenue_it = ₀ + ₁*Rank_itn + ₂*PredLogArrests_it + ₃*LagPromote_it + ₄*LagRelegate_it + ₅*Derby1_itn + ₆*Derby2_itn + ₇*Allseater_it + *Club_i + *Season_t + *Division_n + _it

 PredLogArrests_it is the predicted value of log arrests per 1,000 fans per game computed based on a first-stage regression of the log of Arrests_it on all the exogenous variables in the system plus the log value of arrests per 1,000 fans per game lagged one season (LogLagArrests). In the next section, we report estimates from four different functional forms of equation (3) using linear and logged values of Revenue_it and Rank_itn to ensure robustness of the main findings. These functional forms are: (1) Log Revenue, Rank; (2) Log Revenue, Log Rank; (3) Revenue, Rank and (4) Revenue, Log Rank.

[Insert Table One about here]

      The estimates are from two distinct eras, 1984 to 1994 and then 1999 to 2009. Table One contains summary statistics for all included variables split by era. Several interesting facts arise from an inspection of the variable means. First, it appears that average revenue is lower in the second period; however, recall these are real revenue values deflated using average wage information, indicating that average wages increased to a greater extent than did club revenues. Second, average arrests dropped sharply from ’84-’94 to ’99-’09, as did the minimum and maximum values, which validates common wisdom concerning the overall trend in fan violence. Third, the number of stadia that are all-seater increased dramatically from the early to the later period, reflecting the trend toward “gentrification” of English soccer. 

Results and Discussion

      The data set contains observations from two distinct eras, for seasons starting in 1984 to 1994 and 1999 to 2009. The data are split by eras due to availability, specifically arrest data are unavailable for the seasons starting in 1995 through 1998. However, this split facilitates a test of the influence of fan violence on team revenue in a period characterized by high levels of hooliganism and relatively low levels of revenues (1984-1994). We can then compare these test results to a period characterized by reduced hooliganism and continually increasing team revenues (1999-2009). Figure One shows the time pattern of arrests by division from ’84 to ’94, while Figure Two shows the time pattern by division in the later period.

[Insert Figures One and Two about here]

      Starting with Figure One, it is clear that there was a generally downward trend in arrest rates per 1,000 fans through the 1980s and into the early 1990s for all PEF tiers. It is also clear that arrest rates are higher in lower divisions. The downward trend is likely the result of increased policing efforts and increased attention to crowd control issues by the clubs. The difference by division may be a result of greater effort and resources put into crowd control and policing at clubs in higher divisions. Alternatively, greater arrest rates at lower division clubs may be due to a greater tendency toward violence for fans of such clubs. Gentrification is likely to be connected to ticket prices, which rose much faster in the higher divisions. Also, to the extent that fans view hooliganism (which we might more accurately describe as “activities likely to increase the probability of arrest”) and watching the game as substitute activities, the lower quality of soccer in the lower divisions might imply that hooliganism is a more attractive option in the lower divisions.

      Figure Two indicates that the general trends found in the 1980s and early 1990s continued until about the mid-2000s. Specifically, soccer-related arrests show a downward trajectory for all divisions until about 2004, whereupon the trend flattens out. The average arrest rate for all divisions slowly shrunk from 0.240 per 1,000 fans in the 1984/85 season to 0.047 in 2004. One could conclude from the averages shown in the figures that hooliganism in PEF has stabilized at a rate of about one arrest for every 20,000 fans. In addition, the differences across divisions have not changed much over time; clubs in higher divisions appear to have lower arrests rates. Whether this is a result of division-varying crowd control effort or division-specific differences in fan behavior is an empirical question, one we do not address in the present paper.

Era of “The English Disease”: 1984 to 1994

      Research has highlighted the significance of hooliganism as a social problem in 1980s Britain, so much so that hooliganism has been termed the “English Disease.”¹⁶ Table Two presents estimates of first-stage arrest equation using data from this era in the history of English soccer measuring Rank linearly and in logs. As in all other estimates reported in this paper, team, season, and division fixed effects are included. Both specifications indicate that lower rank in a division is associated with higher arrest rates. Thus, it appears that in the era when hooliganism was at its worst, soccer-related arrests can be linked to poor on-field team performance. In that sense we find that hooliganism was a response to the (low) quality of play of the home team in the early era.

       Teams that are promoted into a higher division tend to bring with them higher arrest rates. There is a tendency for promoted teams to struggle in the new division and the increase in arrest rates for promoted teams may reflect low expectations on the part of fans, who then substitute misbehavior for viewing their teams’ weak performances in a higher division. Alternatively, it could be that promoted teams are “importing” higher rates of fan violence from the lower divisions. The division fixed effects (unreported but available from the authors) show that the level of fan violence increases on average as a team drops down between divisions. However, relegated teams do not deliver greater arrest rates, controlling for rank.

[Insert Table Two about here]

      Having a single home match of local rivalry (Derby1) is associated with more arrests than without a derby. However, having more than one such match does not increase arrest rates any further. This suggests that some matches between local rivals may be more important for fan intensity than others. It may be that there is some sort of fan-violence-concentration issue at work. For example, teams in regions with a single, intense local rival will see fans of both teams truly agitated for relatively few derby games (e.g., Newcastle and Sunderland) while if more local rivals exist, the fan agitation is spread more thinly across a number of derby games. Allseater has insignificant coefficients in both specifications, an unsurprising result since all-seater stadia did not become commonplace until later in this early period.

      As expected, LagArrests has a positive and significant coefficient in the arrests equation. This suggests that this variable can serve as a suitable instrument for arrests in the revenue equation although there is no way to test the validity of a single instrument. Some indirect support for the validity of the instrument lies in the result that when we run an OLS reduced form model of revenues in the early era we find that LagArrests has an insignificant coefficient in the revenue equation.

      Results from estimates of the revenue model for the early period are shown in Table Three. Models with log real revenue are preferred to those with revenue due to better goodness of fit as measured by R². As expected, a higher rank in a division is associated with increased real revenue. Teams that are relegated bring with them greater revenues from the higher division as compared to incumbent teams. Having two or more fixtures with local rivalry raises revenues in this early period. Note that in this early period broadcast revenues were generally a smaller share of total revenues than gate receipts, so revenues are driven primarily by gate attendances, which in turn would be stimulated by having more games of intense, local rivalry.

[Insert Table Three about here]

      In all four specifications in Table Three, the coefficient of our focus variable, PredLogArrests, is negative and significant at the 10% level or better, actually 5% or better in our preferred log revenue form. The finding of a negative relationship between arrests and revenues is robust across several functional forms. The elasticity of revenue with respect to arrests is -0.2 in column (1), which appears to be a plausible value. Both endogeneity tests reject the null hypothesis of exogeneity for the specification in column (1), which we take as our preferred specification. Hence, in the 1980s and 1990s we have new evidence suggesting that increased hooliganism had adverse effects on teams’ financial performances through revenues. The interesting question that follows is whether these adverse effects persist in the more recent era, after the Premier League was formed.

Era of Gentrification: 1999 to 2009

      The top division in PEF broke away to form the English Premier League (EPL) in 1992. One of the primary motivators for this move was so that EPL teams could have independence from the EFL when negotiating television and sponsorship contracts. The mid to late 1990s saw a huge increase in revenues for English soccer, with the lion’s share accruing to EPL clubs. However, revenues for all PEF clubs show a general increase after the mid-90s, owing much to the gentrification of the sport.

      Table Four presents results of both the arrests and revenue equations for the 1999/00 to 2009/10 seasons. As our estimates of arrests and revenues do not vary substantially across functional form for the period, we report results for a single functional form for brevity. The arrests equation now looks very different in the more recent period. The results in the first column (1) of Table Four show little behavioral content in the determination of arrests. The negative relationship between divisional rank and arrests disappears completely, and lagged arrests do not affect current season arrests.

      All we learn from the arrests model is that promoted teams have more arrests and the third and fourth divisions (currently branded as League 1 and League 2) have greater arrest rates than the top two divisions. This latter result was obtained by Green and Simmons (2014) in their analysis of EFL arrests. Green and Simmons argued that English football hooliganism had been reduced over time, as indicated by arrest rates, but that such hooliganism as remained had been displaced to lower division clubs. These clubs tend to be smaller than higher division clubs and lack the necessary financial resources to equip stadiums with better crowd monitoring technology, such as close-circuit television. But the coefficients on lower division dummies and on LagPromote are the only ones to be significant in the arrests model for the more recent era.

[Insert Table Four about here]

      Since lagged arrests do not affect current arrests in the more recent era, the former cannot serve as a valid instrument. It is no surprise that the endogeneity tests in the revenue model in Table Three column (2) comprehensively fail to reject the null hypothesis of exogeneity. The coefficient of PredLogArrests is clearly not statistically significant in the revenue equation. If we replace PredLogArrests by Log Arrests then the coefficient on that variable is also insignificant. Moreover, the revenue results in column (2) appear implausible in that Rank does not have a significant effect on revenues. To reassure us that a plausible revenue model can actually be constructed for the 1999-2009 period we drop arrests altogether in column (3) and the significance of Rank for revenues is then re-established, in line with Szymanski and Smith (1997). 

      We conclude that arrests are orthogonal to club revenues from 1999 and so, in strong contrast to the earlier period, hooliganism has not been detrimental to clubs’ financial performance in England and Wales. Furthermore, team performance within a given division in the period of gentrification does not appear to influence the level of fan violence, which is a clear difference from the earlier period when hooliganism was at its zenith in Britain. There still appears to be a difference among divisions in the period of gentrification, with lower divisions having higher rates of fan violence and promoted teams bringing that violence with them. 

Conclusion

      Soccer hooliganism, defined as episodes of crowd trouble inside and outside soccer stadiums on match days, is commonly perceived to have adverse effects on the sport. However, hooliganism might also have beneficial effects for the home team, if it generates social pressure on players and officials through intimidation. Surprisingly, there appears to be no research estimating the size of these effects (good or bad).

      Identifying these effects empirically is difficult. In this paper, we measure hooliganism by arrests for soccer-related offenses. Using a unique database we find evidence that in the period 1984-94 arrests were positively related to team performance but adversely affected soccer club revenues in England and Wales. This effect disappears in the more recent 1999 to 2009 period showing that hooliganism, still present but at lower levels, no longer has adverse effects on club finances.

      The decline in football hooliganism since the 1980s is consistent with an economic model of crime (Becker, 1964; Ehrlich, 1973; Fajnzylber et al., 2002). Although hooliganism is similar to violent crime in that it can often be instinctive or pathological, it is still conducive to economic analysis. In particular, the opportunity costs of committing hooligan offences have risen through increased detection probability (CCTV; improved stewarding and policing) and greater sentencing costs (banning orders). Also, it is likely that tastes for hooligan activity have shifted with a more diverse (‘gentrified’) and pacific crowd inside the typical English football stadium. While we lack the detailed data needed to disentangle these effects, we have shown that such hooligan activity as remains in England and Wales no longer generates improved team performance and is no longer detrimental to club finances.  
 
 
 

 

References 

Andreff, W. (2014). Building blocks for a disequilibrium model of a European sports league, International Journal of Sport Finance, in press.

Becker, G. (1968) Crime and punishment: An economic approach. Journal of Political Economy. 76: 175-209.

Beech, J., Horsman, S. & Magraw, J. (2008) The circumstances in which English football clubs become insolvent, Coventry University Centre for International Business of Sport Discussion Paper 4.

Dobson S. and Goddard J. (2011) The Economics of Football (2e). Cambridge: Cambridge University Press

Dunning, E (2000) Towards a sociological understanding of football hooliganism as a world phenomenon. European Journal of Criminal Policy and Research, 8,141-162.

Dunning, E., Murphy P. & Williams J (1988) The Roots of Football Hooliganism: a Historical and Sociological Study. London: Routledge and Kegan Paul.

Ehrlich, I. (1973). Participation in illegal activities: a theoretical and empirical investigation. Journal of Political Economy, 81, 551-565.

Fajnzylber, P., Lederman, P. and Loayza, N. (2002). What determines violent crime? European Economic Review, 46, 1323-1357.

Garicano L., Palacios-Huerta I, & Prendergast C (2005) Favoritism under social pressure. Review of Economics and Statistics, 87, 208-216.

Green, C. and Simmons, R. 2014. The English disease: Has football hooliganism been eliminated or merely displaced? Lancaster University Management School Working Paper.

Harrington, J. (1968). Soccer Hooliganism. Bristol: John Wright

Marsh P, Rosser E & Harré R. (1978). The Rules of Disorder. London: Routledge & Kegan Paul.

Pearson, G. (1999). Legitimate targets? The civil liberties of football fans. Journal of Civil Liberties, 4, 28-47.

Pettersson-Lidborn, P. and Priks, M. (2010). Behavior under social pressure: Empty Italian stadiums and referee bias. Economics Letters, 108, 212-214.

Poutvaara, P. & Priks, M. (2009). Supporter violence and police tactics. Journal of Public Economic Theory, 11, 441-453.

Priks, M. (2010). Does frustration lead to unruly behaviour? Evidence from the Swedish hooligan scene. Kyklos, 63, 450-460.

Priks, M. (forthcoming). Do surveillance cameras affect unruly behaviour? A close look at grandstands. Scandinavian Journal of Economics.

Rees, D & Schnepel, K. (2009). College football games and crime. Journal of Sports Economics, 10, 68-87.

Rookwood, J. and Pearson, G. (2012) “The hoolifan: Positive fan attitudes to football ‘hooliganism’ ”, International Review for the Sociology of Sport, 47(2), 149-164

Stott, C. & Pearson, G. (2007). Football ‘hooliganism’. London: Pennant Books.

Szymanski, S. (2012). Insolvency in professional football: Irrational exuberance or negative shocks? International Association of Sports Economists and North American Association of Sports Economists, Working Paper 1202.

Szymanski, S. & Kuypers, T. (1999). Winners and losers: The business of football. London: Viking Press.

Szymanski, S. & Smith, R. (1997). The English football industry: Profit, performance and industrial structure. International Review of Applied Economics, 11, 135-153.

Taylor I. (1971). Football mad: a speculative sociology of football hooliganism. In E Dunning (ed.) The Sociology of Sport: A Selection of Readings. London: Frank Cass.

Lord Justice Taylor (1990) The Hillsborough Stadium Disaster. London: HMSO, Cm962

Taylor, M. (2008) The Association Game: A History of British Football. Pearson.

Tsoukala, A. (2009)  Football Hooliganism in Europe: Security and Civil Liberties in the Balance. London: Palgrave Macmillan 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Table One

Summary Statistics 

1984 to 1994 (n = 821) 

Variable Mean Std. Dev. Min  Max

Revenue 48.08        71.30 0.24 420.40

LogRevenue 2.85 1.46 -1.43 6.04

Arrests 0.15 0.14 0.0058 1.14

LogArrests -2.21 0.85 -5.15 0.13

LagLogArrests -2.12     0.87   -5.94   0.24

Rank 11.85 6.63 1 24

LogRank -2.23 0.81 -3.18 0

LagPromote 0.11 0.31 0 1

LagRelegate 0.11 0.31 0 1

Allseater 0.09 0.29 0 1

Derby1 0.29 0.45 0 1

Derby2 0.08 0.27 0 1 

1999 to 2009 (n = 764) 

Variable Mean Std. Dev. Min  Max

Revenue 40.20        59.63 0.17 379.86

LogRevenue 2.81 1.34 -1.78 5.94

Arrests 0.05 0.04 0.0014 0.32

LogArrests -3.16 0.77 -6.57 -1.14

LagLogArrests -3.09     0.78   -6.57  -0.84

Rank 11.45 6.65 1 24

LogRank -2.18 0.83 -3.18 0

LagPromote 0.11 0.32 0 1

LagRelegate 0.11 0.32 0 1

Allseater 0.79 0.41 0 1

Derby1 0.26 0.44 0 1

Derby2 0.09 0.29 0 1 
 
 
 
 
 
 

Table Two

Arrests Equation

1984 to 1994 (n = 821) 

Variable	Coefficient (t statistic)	Coefficient (t statistic)
LagLogArrests	0.159 (4.41)***	0.159 (4.39)***
Rank	-0.011 (2.67)***
Log Rank		-0.073 (2.18)**
LagPromote	0.267 (3.30)***	0.259 (3.20)***
LagRelegate	0.021 (0.26)	0.023 (0.28)
Allseater	-0.101 (0.92)	-0.103 (0.93)
Derby1	0.131 (1.92)*	0.133 (1.96)**
Derby2	0.144 (1.10)	0.148 (1.12)
Adjusted  R2	0.434	0.432

*significant at 10% level

**significant at 5% level

***significant at 1% level 
 
 

 

Table Three

Revenue Equation

1984 to 1994 (n = 821) 

Dependent Variable	Log Revenue	Log Revenue	Revenue	Revenue
	Coefficient (z statistic)	Coefficient (z statistic)	Coefficient (z statistic)	Coefficient (z statistic)
PredLogArrests	-0.198 (2.01)**	-0.187 (1.91)*	-15.33 (2.11)**	-15.42 (2.12)**
Rank	0.014 (6.27)***		0.307 (1.92)*
Log Rank		0.117 (7.09)***		1.733 (1.41)
LagPromote	0.052 (1.22)	0.054 (1.31)	1.776 (0.57)	2.086 (0.68)
LagRelegate	0.105 (2.97)***	0.103 (2.95)***	1.353 (0.52)	1.310 (0.50)
Allseater	0.059 (1.19)	0.067 (1.37)	0.201 (0.06)	0.191 (0.05)
Derby1	0.021 (0.65)	0.020 (0.62)	3.280 (1.36)	3.185 (1.31)
Derby2	0.125 (2.09)**	0.120 (2.02)**	3.439 (0.78)	3.334 (0.76)
R2	0.963	0.964	0.917	0.917
Endogeneity Tests (p values)
Durbin	0.068	0.092	0.070	0.073
Wu-Hausman	0.090	0.118	0.092	0.095

*significant at 10% level

**significant at 5% level

***significant at 1% level 

 

Table Four

Arrest and Revenue Equations

1999 to 2009 (n = 764) 

Dependent Variable	Log Arrests First Stage	Log Revenue 2SLS	Log Revenue OLS
	Coefficient (t statistic)	Coefficient (z statistic)	Coefficient (t statistic)
PredLogArrests		-1.132 (0.48)
LogLagArrests	0.017 (0.42)
Rank	0.0003 (0.08)	0.009 (1.08)	0.012 (6.42)***
LagPromote	0.240 (3.40)***	0.331 (0.45)	-0.014 (0.42)
LagRelegate	-0.067 (0.97)	0.104 (0.62)	0.176 (5.27)***
Allseater	-0.099 (0.61)	0.102 (0.43)	0.178 (2.53)**
Derby1	0.026 (0.44)	0.081 (0.94)	0.055 (1.86)*
Derby2	0.091 (0.98)	0.196 (0.64)	0.060 (1.29)
Division2	0.094 (1.18)	-1.019 (2.95)***	-1.179 (30.01)***
Division3	0.297 (2.52)**	-1.271(1.41)	-1.697 (31.64)***
Division4	0.333 (2.28)**	-1.678 (1.45)	-2.223 (32.45)***
R2	0.543	0.809	0.963
Endogeneity Tests (p values)
Durbin		0.25
Wu-Hausman		0.29

*significant at 10% level

**significant at 5% level

***significant at 1% level

 

Figure One

Arrest Rate per 1,000 Fans per Game

1984 to 1994

By PEF Division 

 

Figure Two

Arrest Rate per 1,000 Fans per Game

1999 to 2009

By PEF Division 
 

¹ As early as 1968, The Harrington Report warned that hooliganism was affecting attendance levels (chapter 11, p. 52), and recent works such as Dobson and Goddard (2011, p. 156) confirm this belief.

 

² Priks (forthcoming) offers evidence from Swedish micro data to show that the introduction of CCTV can causally lead to a reduction in disorderly behavior.

³ By “gentrification” we mean football crowds that are composed of older fans with higher incomes. This may reflect a change in the social composition of the fans (a larger share of white-collar, professional, technical or managerial employees) or alternatively just an aging of the fan base, and hence a lower propensity to rowdy behavior.

⁴ An excellent survey of this literature is provided by Taylor (2008, pp. 309-319).

 

⁵ The Hillsborough disaster is in many ways the defining event of modern English football and the watershed. It has taken many years to establish that some in the police authorities badly mismanaged the response the disaster and then attempted to cover up the evidence. Ultimately, however, there has been a significant re-evaluation of the relationship between fans and the police as a result of what happened. The events and subsequent response to Hillsborough has been exhaustively documented on the website of the Independent Panel hillsborough.independent.gov.uk. Tsoukala (2009) assesses the relationship between policing and hooliganism across Europe from a criminologist’s perspective.

⁶ The North American literature on sports-related offences is sparse. Rees and Schnepel (2009) argue that arrest rates in college (American) football towns rise when the home team suffers an unexpected loss.

⁷ Hooliganism could be one factor that enhances home advantage. Referees might offer greater decision-making bias in favor of home teams if they are successfully intimidated by potentially violent home fans. In Italy, stadia that are scenes of severe outbreaks of hooliganism (including racist behavior) are closed to all fans for a number of subsequent games. Pettersson-Lidbom and Priks (2010) present evidence on Italian football to show that referee bias was reduced when home teams were forced to play behind closed doors, i.e. without fans present.

⁸ Hooliganism could be a manifestation of fan frustration with team performance relative to what might be expected (Priks, 2010).

⁹ This might be a direct consequence of disappointment at defeat, but it might also be a signal of dissatisfaction with club management, especially the Chairman and Board of Directors, and thus a form of lobbying for reform.

¹⁰ Financial information for the last decade or so at least can downloaded for a small fee from this website http://wck2.companieshouse.gov.uk/c3e6ddbd60434f4088c2fcfcf1db83ac/wcframe?name=accessCompanyInfo. Data from 1974 to around 2000 can be obtained by request on a CD-rom. Prior to 1974,only paper records exist.

¹¹ We could have left revenues in nominal terms, letting year dummies take care of inflation, or we could deflate revenues by CPI letting year dummies control for ‘excess’ inflation. Our results are robust to these alternative treatments of revenue.

¹² “Arrests” covers all arrests defined in law under schedule 1 of the Football Spectators Act 1989 and includes football specific offences such as violent behavior, throwing missiles and pitch encroachment and a wide range of generic criminal offences committed in connection with a football match. Note that the offences need not be committed immediately before, during or after the match; the Home Office states that a football-related arrest can occur “at any place within a period of 24 hours either side of a match.” This flexibility is designed to capture alcohol-related offences that might occur sometime before or after matches are actually played.

¹³  We tested the “frustration hypothesis” of Priks (2010) by running a first-stage OLS regression model of team ranks on team payrolls. We then used the residuals from this regression as an instrument for arrests in our revenue-arrests model. The coefficient on this instrument was insignificant so we could not find support for the frustration hypothesis, at least by this approach.

¹⁴ Using a single exclusion restriction implies that the fan violence equation is exactly identified. We investigated additional exclusion restrictions, including measures of policing intensity, without finding any other suitable instruments. The major drawback to our use of a single exclusion restriction is that there is no way to statistically test for the appropriateness and strength of the exclusion restriction in exactly-identified equation systems.

¹⁵ We estimated models that included measures of whether a club went into insolvency and the club’s debt-to-assets ratio. However, no significance was found in these measures in any subset of the data. All estimates are available from the authors. One reason why insolvency might not affect revenues or arrests is the presence of a soft budget constraint (Andreff, 2014). Insolvent soccer clubs are not usually wound up so fans do not necessarily care about bad financial performances of their clubs so long as their teams can continue to play (i.e., fulfill their fixtures). A similar argument applies to the debt-assets ratio. We expect that when teams underperform in the league then their revenues fall relative to spending. The debt-assets ratio rises and if this is not controlled then insolvency is the result. In that sense, insolvency is a consequence and not a driver of low revenues. Also, hooliganism (arrests) will tend to be orthogonal to the vicious circle of debt and insolvency that has plagued several English lower division clubs, as documented by Beech et al.(2008). Szymanski (2012) articulates a model of debt and insolvency in English soccer.

 

¹⁶ “The English disease” is a term that was used by several journalists to characterize crowd trouble triggered by English football hooliganism both at home and abroad.