Content uploaded by Justin Kurland
Author content
All content in this area was uploaded by Justin Kurland on Jul 26, 2019
Content may be subject to copyright.
An analysis of the Spatio-temporal ‘footprint’ of crime in and
around Sheffield Wednesday FC (Hillsborough)
August 2011
Address for correspondence:
Professor Nick Tilley
Department of Security and Crime Science
University College London
35 Tavistock Square
London
WC1H 9EZ
Tel +44 (0)20 3108 3160
Fax +44 (0)20 3108 3088
Email: n.tilley@ucl.ac.uk
UCL DEPARTMENT OF SECURITY AND CRIME SCIENCE
An analysis of the Spatio-temporal ‘footprint’ of crime in and
around Sheffield Wednesday FC (Hillsborough)
Contents
Executive summary...................................................................................... iv
Acknowledgements...................................................................................... v
Report
1. Introduction........................................................................................ 1
2. Background........................................................................................ 1
3. Data and Methods.............................................................................. 2
4. Results................................................................................................ 5
5. Conclusions and Recommendations.................................................. 11
Appendix One: Formalities of Football policing and GMP v. Wigan AFC. 13
Appendix Two: List of Sheffield Wednesday home matches and their
comparators.................................................................................................... 15
Appendix Three: Technical Overview of Statistical Methods..................... 16
Appendix Four: Spatial Permutation Significance Table............................ 18
Appendix Five: Temporal Permutation Significance Table......................... 19
Appendix Six: Detailed Report of Site Visits............................................... 20
References..................................................................................................... 22
iii
List of Tables
Table 4.1: Count and Ratio of match-day crimes and incidents to non-match day
crimes and incidents in the study area
Appendix Two: Football Match and Comparison Days
Appendix Four: Spatial Permutation Significance Result Table
Appendix Five: Temporal Permutation Significance Result Table
List of Figures
Figure 3.1: Map of Sheffield Wednesday study area
Figure 4.1: Match and comparison day ratio of crime counts by distance from
Hillsborough
Figure 4.2: Temporal pattern of crimes within the 1500 metre area around
Hillsborough on match and comparison days
Figure 4.3: Temporal pattern of incidents within the 500 metre area around
Hillsborough on match and comparison days
Figure 4.4: Temporal pattern of violence against the person offences within the
500 metre area around Hillsborough on match and comparison days
Figure 4.5: Temporal pattern of theft and handling offences within the 750 metre
area around Hillsborough on match and comparison days
iv
Executive Summary
The main objective of this study, commissioned by the South Yorkshire Police
(SYP), the Association of Chief Police Officers (ACPO) and the UK Football
Policing Unit (UKFPU), is to assess whether the current geographic ‘footprint’
within which football clubs must pay for policing services on match days
sufficiently covers the costs of match policing in areas in and around six
professional football stadia in England. The current report focuses on one of
these, addressing the above question for Sheffield Wednesday (hereafter
Hillsborough).
Toward this end, this study employs six years (2005-2010) of geo-coded case-
level data supplied by SYP captured the crime and incidents occurring within
three kilometres of Hillsborough. These data cover five football seasons,
beginning with 2005-6. All 126 dates when Sheffield Wednesday played home
matches at Hillsborough during this period were identified. A Java-based
program isolated 98 suitable comparator dates when matches did not take place. A
total of 5,775 crimes and 20,294 incidents for match and non-match days were
used in these analyses.
The data for these dates, the stadium and the rail stations falling within the
catchment area were electronically mapped via ArcGIS 9 to discern any spatial
patterns. The recorded crime and incident patterns associated with match and non-
match days were then statistically compared.
The findings suggest a heightened number of recorded crimes on match days in
areas extending up to 1500 metres away from the centre of the ground. Two
caveats need to be entered in interpreting this finding. First, the presence of the
police may inhibit some criminal and antisocial behaviour, although the degree to
which this is the case could not be ascertained. Second, casual observation of
crowd and police behaviour on match days and anecdotal evidence from the police
suggest that at Hillsborough (as at other grounds) police recording practices on
match days differ from those on non-match days: pressure of work in marshalling
fans and more tolerance for low-level antisocial behaviour in accordance with
distinctive match-day norms mean that what might be recorded on a non-match
day is typically not recorded on match-days.
The spread of elevated levels of recorded crime and antisocial behaviour around
Hillsborough is as great as or greater than that observed at other football stadia in
the study. Hillsborough is just north of the city centre and geographically isolated
from public houses. The lack of pubs in the local area causes both home and away
supporters to find establishments at farther distances from the ground then at other
stadia. These geographical factors may facilitate increased crime in areas away
from Hillsborough.
Future research might test more directly hypotheses relating to the impact of
policing practices and geography on recorded patterns of crime and antisocial
behaviour.
v
Acknowledgements
We are delighted to acknowledge help from a range of sources in conducting the
research reported in this document.
We are grateful for support provided by South Yorkshire Police (SYP).
Superintendent Rachel Barber of the South Yorkshire Police was invaluable in
giving direction to those who could provide us with the necessary data.
Superintendent Barber, at very short notice, arranged a date for a site visit to
Hillsborough to observe the policing operation on a football match day.
A major problem for research of this type relates to the supply and availability of
data. We are indebted to a range of exceptionally helpful individuals, who
responded positively and quickly to our requests for data. They went to great
trouble in extracting and sanitizing the data so that they could be utilised for the
purposes of this project. Those we thank include: Geo-coding clerk Joseph
Williams, Detective Chief Inspector Gary Williams and Detective Chief Inspector
Steve Williams of the SYP.
Department of Security and Crime Science UCL
1
1. Introduction
This research project was commissioned by the South Yorkshire Police (SYP), British
Transport Police (BTP), the Association of Chief Police Officers (ACPO) and the UK
Football Policing Unit (UKFPU) in January 2011. These agencies were interested in
the spatial and temporal distribution of crime and disorder incidents around football
stadia on match days. As such, a meeting took place between the ACPO Football
Lead ACC Andy Holt, UKFPU Director Bryan Drew and members of the Department
of Security and Crime Science at UCL. The meeting focused on how previous
research conducted by Professor Nick Tilley, Dr Shane D. Johnson and Justin
Kurland for the Government Office for London uncovered elevated crime and
disorder patterns on football match days as compared to non-match days and how the
research could be replicated and extended to other football stadia in England.
Resulting from this meeting, the Department of Security and Crime Science at UCL
was invited to undertake the research. Their findings for one of these stadia,
Hillsborough Stadium, home of Sheffield Wednesday, are presented below. The next
section gives a brief overview of the football ‘footprint’ and existing research that
gave rise to the present analyses. It is followed by a description of the methods and
data used. The resultant findings are then presented. The final section draws
conclusions from the quantitative and qualitative results, providing recommendations
for further research and practice.
2. Background
Football clubs often depend on police support to ensure the safety of matches as well
as the individuals drawn to the area by them. Indeed, clubs must obtain a safety
certificate from the local authority who are required to determine the safe capacity of
the ground and to prescribe and enforce such terms and conditions as it considers
necessary to secure the safety of spectators. . As such, cooperation between the clubs
and the police is paramount to public order and safety as well as to the sport itself.
However, payment arrangements can lead to tension. During the 2003/2004 season
the Greater Manchester Police (GMP) proposed to increase the charges for police
services to Wigan AFC after the promotion of that club to Division One (the
Championship) from Division Two (League One). Official statistics suggested that
this success would substantially increase match attendance at Wigan AFC in the
coming season, thereby necessitating augmented policing levels on match days and
justifying increased charges. Wigan AFC did not accept this and compensation for
policing services at Wigan went unpaid for the duration of the 2003/2004 season,
despite the on-going provision of services by GMP. This continued into the following
year, even though the Football Licensing Authority (FLA) pressured Wigan AFC to
make payment. Though Wigan AFC eventually made some payment, GMP
ultimately sued in court for the remainder (£293,085). The Courts initially ruled in
favour of GMP, but this judgement was reversed on appeal.
2
Importantly, the GMP v Wigan AFC decisions produced clear guidelines on when and
where it is appropriate to charge for football match policing, how much can be
charged and the means by which decisions on the level of policing services are
rendered (for a more detailed explanation of the judgement and current procedures
please refer to Appendix One). These guidelines also deal with the spatial deployment
of police at football matches and specify the areas within which clubs are required to
pay for police services. Collectively, these are commonly referred to as the football
‘footprint’. However, GMP v Wigan AFC presented no empirical evidence
demonstrating the appropriateness (or otherwise) of existing ‘footprints’.
3. Data and Methods
Previous research (Kurland, Johnson, and Tilley 2010) drew on two years (March
2007-November 2009) of geo-coded crime and incident data from the Metropolitan
Police Service (MPS) to assess the spatial and temporal patterns of crime within one
mile of Wembley Stadium. Comparing these rates for football match days with those
for equivalent non-match days, their findings indicate that match days have
significantly (p<0.001) higher levels of crime and disorder up to 750 metres from the
stadium. Such results suggest that clubs may not be adequately held responsible for
the costs associated with policing football matches, since the typical ‘footprint’ of
approximately 100m is much smaller than 750 metres.
The current study reports the analysis of geo-coded crime and incident data provided
by SYP. The study made use of 5,775 geo-coded crimes and 20,294 geo-coded
incidents provided by SYP1. Replicating and extending Kurland et al (2010) for
Hillsborough, it employs a three kilometre radius and six-year timeframe (2005
through 2010). This aims to provide a more thorough estimation of how spatial and
temporal patterns of crime and incidents may be differentially associated with match
and non-match days.
Matches were played at Hillsborough on 126 dates in the 2004/5—2009/10 seasons.
For 98 of these, a comparable non-match day was identified via a Java-based
programme2. This paired-date approach makes it possible to determine whether the
1 The SYP provided data set for crimes included a total of 86,456 geo-coded crimes and a total of
311,550 geo-coded incidents.
2 The programme identifies the seven days before and the seven days after a given match, selecting one
2 The programme identifies the seven days before and the seven days after a given match, selecting one
optimal comparator date in each week. If no match occurs on either date for Sheffield Wednesday, or
Sheffield United, and if neither has already been selected as the comparison day for another match, the
earlier date is selected as the match comparator to leave sufficient comparison days for matches
occurring later in the fixture list. However, if both dates are unavailable due to matches or prior
selection, then the programme search parameters are first expanded to the fourteen days before and
fourteen days after the match, and if no match is found then twenty-one days and if not match is found
again then twenty-eight days. If no suitable comparison days were found within this twenty-eight day
period from the match day that particular match was excluded from the sample match day data set.
Ultimately, a total of sixty-three comparison days fell within seven days of the match, fourteen
comparison days fell within the fourteen day period, thirteen within twenty-one days whilst eight were
within twenty-eight days.
3
patterns of crime and incidents in and around Hillsborough vary between match and
non-match days. Appendix Two provides a full listing of the match and comparator
dates.
For ease of analysis and reporting, the crime data were cleansed and disaggregated by
the Home Office categories of recorded crime,3 yielding three crime-specific groups,
one amalgamated category for crime types too small for substantive analysis
individually and one general crime grouping that combined all Home Office
categories. In total, there were 5,775 usable crimes and 20,294 usable incidents
occurring on match and non-match days respectively4.
Crimes were geo-coded and assigned to radiating buffers of 10 metres extending
cumulatively to three kilometres from the centre of the ground. Figure 3.1 is a map
illustrating the buffers around the stadium, identifying the city centre as well as the
main transport nodes. Statistical tests were used to examine the reliability of spatial
patterns and differences in the counts of events at varying buffer sizes for match and
comparison days.
Next the events that took place within the spatial areas identified as being
significantly different on match days were re-categorised according to the time of day
that they occurred. The day was divided into six specific hourly intervals: 0 to 3, 4 to
7, 8 to 11, 12 to 16, 17 to 20 and 21 to 24. A statistical test was then used to examine
the differences in counts for these temporal intervals for match and comparison days.
In addition, Justin Kurland undertook a match-day site visit to Hillsborough. He
attended police briefings, observed match day dynamics and informally interviewed
the personnel participating in policing that match. This provided insight into the
police-perceived match-day ‘hotspots’ (or ‘flash-points’) as well as officer views of
their primary match-day roles.
3 These are: Burglary, Criminal Damage, Drug Offences, Fraud and Forgery, Robbery, Sexual
Offences, Theft and Handling, Violence against the Person and Other Notifiable Offences. Although
varied recording practice is a common problem with police-recorded data, the April 2002 National
Crime Recording Standard minimises this issue (Simmons, Legg, and Hosking 2003)
4 There were a total of 80,681 unusable crimes and 291,256 unusable incidents. The majority of these
were not utilised because they were crimes and incidents that did not take place on either match or
comparison days. Others crime and incident geo-coded data were excluded if they fell outside of the 3
kilometre radius of the football ground.
4
Figure 3.1
5
4. Results
The first question that the findings reported here address is, ‘Over what area
surrounding the stadium, if any, have increased numbers of police recorded crimes or
incidents been associated with days when football matches were played at
Hillsborough?’ The second is ‘Within any area of increased numbers of recorded
crimes or incidents over what time during the day have the increased numbers
occurred?’
a) Over what area surrounding the stadium, if any, have increased numbers of
police recorded crimes or incidents been associated with days when football
matches were played at Hillsborough?
On average there were more total crime and incidents that took place on match days
then non-match comparison days. Overall, there were a total of 3,097 match day
crimes and 2,678 non-match day crimes as well as a total of 10,538 and 9,756
incidents, respectively. Presenting only amalgamated totals for crime and incidents
would mask their variation across space and so a table with spatially disaggregated
counts was constructed for several categories and are presented in Table 4.1. The top
segment of the table shows the counts of crime and incidents for match days by
radiating non-overlapping 250 metre buffers, the middle segment does the same for
non-match days, and the bottom one shows the ratio of match to non-match day
counts. The top segment also shows the number of hectares in each buffer, to remind
the reader that radiating buffers of a constant width increase in area as they grow
more distant from the centre. For example, the first buffer with a radius of 250
metres is only 20 hectares, whilst the final buffer (2750 to 3000 metres) covers an
area of 452 hectares.
The third row in the top segment of Table 4.1 shows that there were 261 incidents in
the first buffer on match days. The equivalent cell in the middle segment shows that
there were 101 incidents for the same area on the comparison non-match days. Thus,
relative to non-match days, there were (261/101=) 2.6 times as many incidents
recorded by the police on match days at locations that were within a radius of 250
metres from the centre of the ground.
Overall Table 4.1 shows there were 3.1 times as many recorded crimes in the first
buffer, including 3.6 times as many cases of theft/handling (although the counts here
are very small at 43 and 12 respectively for match and non-match days) and 6.2 times
as many violent crimes (37 and 6 on match and non-match days). For the remaining
crime categories shown in the table there was little difference or there were fewer
recorded crimes. The elevated differences between match and non-match days persist,
albeit to a lesser extent, in the area between 250 to 1,500 metres for total crime, 250
and 1500 for violence against the person offences, 250 and 1250 for theft and
handling and 250 and 1500 for the amalgamated other crime category. Moving
beyond these identified areas the data indicate that differences between match and
non-match days exist but outside these areas differences are generally small, if any,
and have no consistent pattern.
6
Table 4.1
Sheffield Wednesday
Counts for match days
Buffer
0-
250
250-
500
500-
750
750-1000
1000-
1250
1250-
1500
1500-
1750
1750-
2000
2000-
2250
2250-
2500
2500-
2750
2750-
3000
Area (hectares)
20
59
98
137
177
216
255
295
334
373
412
452
Incidents
261
343
394
770
927
898
891
1193
985
856
1424
1596
Crimes
96
121
120
294
327
320
212
313
271
223
337
463
Violence
37
24
20
74
52
57
48
47
52
32
77
72
Theft/handling
43
43
61
97
126
96
77
102
76
81
88
163
Criminal damage
5
20
21
67
95
102
56
93
78
66
92
136
Other
11
34
18
56
54
65
31
71
65
44
80
92
Counts for comparison days
Incidents
101
234
402
667
842
881
837
1117
869
943
1299
1564
Crimes
31
65
88
210
280
258
216
295
236
251
330
418
Violence
6
7
18
42
53
44
48
51
53
44
69
62
Theft/handling
12
22
35
89
100
93
80
94
69
64
95
166
Criminal damage
6
15
26
47
79
64
49
85
69
76
81
101
Other
6
21
9
32
48
57
39
65
45
67
85
89
Ratio of match to comparison days
Incidents
2.6
1.5
1.0
1.2
1.1
1.0
1.1
1.1
1.1
0.9
1.1
1.0
Crimes
3.1
1.9
1.4
1.4
1.2
1.2
1.0
1.1
1.1
0.9
1.0
1.1
Violence
6.2
3.4
1.1
1.8
1.0
1.3
1.0
0.9
1.0
0.7
1.1
1.2
Theft/handling
3.6
2.0
1.7
1.1
1.3
1.0
1.0
1.1
1.1
1.3
0.9
1.0
Criminal damage
0.8
1.3
0.8
1.4
1.2
1.6
1.1
1.1
1.1
0.9
1.1
1.3
Other
1.8
1.6
2.0
1.8
1.1
1.1
0.8
1.1
1.4
0.7
0.9
1.0
Note: ‘Other’ includes crime types too small in number for meaningful individual analysis. It combines burglary, sex crimes, robbery, fraud/forgery, drugs crimes, and other.
A finer resolution of analysis is shown as Figure 4.1, this time for cumulative
geographic buffers. As before, the contrast is the ratio of all crimes on match days to
those on non-match comparison days. In this case, the entire study area was divided
into 300 concentric ring shaped buffers, each 10 metres wide, with a circular buffer
with a ten metre radius in the middle (shown on the horizontal, X axis). Then starting
with the innermost buffer, and radiating out to the full distance of 3000 metres, counts
of crimes were made for match and non-match days to calculate the spatial crime
ratios (shown on the vertical, Y axis). Thus, farthest to the left the graph shows the
ratio of match day crimes to comparison day crimes for the 12 innermost circles (with
a 120 metre radius =51/3, or 17:1) and farthest to the right the graph shown the ratio
for the full set of incidents in the analysis in the complete circle round the ground
(with a 3000 metre radius = 3084/2667, or 1.2:1). This graph suggests that there is a
steep decline in the relative difference between match and non-match days in the
count of incidents as the distance from the ground increases. By 250 metres the ratio
is 3.1 to one, and by 500 metres it reduces to 2.3 to one.
Figure 4.1
Were the differences between match and non-match days statistically significant or
could they be explained by random variation? Conventionally, if the probability that
an observed effect would be observed on a chance basis less than one time in twenty
(i.e. less than 5% of the time, or in more formal terms if p<0.05) then the outcome is
accepted as being statistically significant. Using this standard, and two different
types of non-parametric permutation tests (see Appendix 3 for a description), we
found that the number of crimes between the two sets of days within the 1500 metre
area differed significantly (p<0.003). For simplicity and because the results were
generally similar we report the results of only one test. In addition, further analysis
indicated that within the 1500m buffer, the spatial distribution of crimes differed
0500 1000 1500 2000 2500 3000
0 5 10 15 20
Distance buffer starts from Stadium Centre(m)
Ratio
Match to Comparison Index
All C ri me s
8
significantly (D= 0.1867, p<0.005) for match and non-match days (see Appendix 3
for a description).
Using the same permutation tests we found that the number of violence events and
incidents between the two sets of days within the 500-metre area also differs
significantly (p’s<0.05). Additionally the number of theft and handling and other
events between the sets of days differed significantly (p’s<0.05) within an even wider
area of 750 and 1000 metres respectively.
The permutation test also identified a ‘hotspot’ where a significantly higher number
of criminal damage events (p<0.05) take place between 750 and 2250 metres away
from the centre of Hillsborough.
b) Within any area of increased numbers of recorded crimes or incidents, over
what time during the day have the increased numbers occurred?
The results reported here on the distinctiveness of the time of day when crimes and
incidents on match days take place relate to recorded crimes across the area over
which overall numbers were found to be statistically significant. For example, 1500
metres from the centre of Hillsborough was examined temporally for all crime but a
smaller area of 500 metres was used for incidents because these were the areas over
which differences in numbers of the respective events were found to be significantly
different. Figure 4.2 shows the numbers of crimes at different times of the day for
match and non-match days. It also shows the range of start and finish times for
matches: the black and grey dotted vertical lines correspond to the respective median
start and finish times. The associated solid horizontal lines indicate the inter quartile
range for each; for instance, 50% of matches started within the range delineated by
the space in between the two black horizontal lines in (i.e. roughly 3pm-7pm). The
earliest start time was noon and the latest 8pm. It can be seen that there are more
crimes on match than non-match days with four distinct peaks occurring at match
start and finish times. Figure 4.3 does the same for incidents.
Was there a statistically significant difference in the volume of crimes or incidents on
match and non-match days at specific times? Again using a non-parametric
permutation test (see Appendix 3 for a description), we found that the number of
crimes between the two sets of days within 1500 metre area over the course of the day
differed significantly (p<0.003) for the period between noon to 23:00 hours. There
were also differences in the volume of incidents that take place on match days from
08:00 to 19:00 across a distance of 500 metres from the centre of Hillsborough
(p’s<0.05). For violence against the person, within the 500m buffer and theft and
handling, within the 750m buffer, the count of events were deemed too low to
conduct these more precise analyses, however Figures 4.4 and 4.5 demonstrate a
pattern of increased counts during match times for each of the respective crime types.
9
Figure 4.2: Temporal pattern of crimes within the 1500 metre area around
Hillsborough on match and comparison days
Figure 4.3: Temporal pattern of incidents within the 500 metre area around
Hillsborough on match and comparison days
10
Figure 4.4: Temporal pattern of violence against the person offences within the 500
metre area around Hillsborough on match and comparison days
Figure 4.5: Temporal pattern of theft and handling offences within the 750 metre
area around Hillsborough on match and comparison days
11
5. Conclusions and Recommendations
The statistical data analysed in this report provide evidence of significant increases on
match compared to non-match days in the area not only immediately surrounding the
ground but in area for certain crime types (criminal damage) that are removed from
the immediate vicinity surrounding Hillsborough. Hence, on their own they provide
evidence to increase the area over for which the club might be asked to contribute to
the costs of policing.
These statistical findings, however, are subject to limitations that will tend to
underestimate the increased levels of crime and disorder associated with match days
at Hillsborough. First, site visit observations suggest that police working on match
days are often too busy to record all observed incidents. Second, given the larger
security implications of their actions, match day officers were observed to disregard
lower-level crime and incidents rather than taking action (e.g. making an arrest) that
would take them away from the scene and the ability to manage any serious disorder
that might break out. Third, the norms of acceptable behaviour used by the police for
those at football matches appear more tolerant of actions that could in principle lead
to formal action than those used by the police in other circumstances. The police,
thus, appear to use their enforcement discretion differently on match days as
compared to non-match days5. These factors suggest a shift in police priorities, from
enforcement to effective order maintenance, when dealing with football matches. This
is consistent with previous research on police decision-making (e.g. Klinger 1997).
Recording practices thus undoubtedly vary between match and non-match days
making the current results a highly conservative estimate of any actual differences in
crime levels. In addition it may also be that the increased police presence in and
around Hillsborough on match days deterred crime and disorder. Specifically and as
shown by the site visit, when problems are expected, the police position themselves
so as to anticipate and prevent crime and disorder.
The findings presented in this report are, of course, subject to the limitations of data
recorded by the police. One such limitation relates to the accuracy with which
incidents are geo-coded. In the best case scenario, each incident would be geo-coded
to the exact coordinates that describe where it took place. However, in practice for
crimes that take place on the street, victims are not always able to provide a complete
description of the offence location, reporting perhaps instead the general point on the
street at which the offense took place. As a consequence, the police are sometimes
only able to estimate the location of where events would have occurred (e.g. by
allocating a crime to a particular street), and this can lead to a loss of precision in the
geo-coding of some incidents. Importantly, however, this type of issue should apply
equally to offences that occur on both match and non-match days, which should limit
the extent to which it could affect the analyses of the spatial distribution of crimes.
That said, the reader should be aware of this issue. Such issues would, of course, not
affect the other forms of analysis.
5 Despite what appears to be happening with discretionary variation this final conclusion remains only
tentative because no non-match day policing observations were undertaken as part of this research.
12
The elevation in levels of recorded crime and antisocial behaviour around
Hillsborough is as great or greater than that observed at other grounds in this study.
The times in which the elevation in these recorded crimes also appears to be directly
related to the times in which football matches take place. This may be due to
geography as well as police action. Hillsborough is south of the city centre and there
are few public houses in close proximity to Hillsborough. The lack of public houses
around Hillsborough forces both home and away supporters who are looking to have
drinks away from the ground prior to match. This causes large numbers of supporters
to walk down various arterial roads on their way to and from the ground thus
increasing the likelihood of confronting opposing supporters or engaging in criminal
damage without the presence of capable guardians or police.
Future research might test more directly test hypotheses relating to the impact of the
geography around Hillsborough on recorded patterns of crime and behaviour.
13
Appendix One: Formalities of Football policing and GMP v. Wigan AFC
GMP v. Wigan AFC revolved around the legal aspects of special police services
provided for professional football matches. The original case and appellate rulings
produced guidance for the police and football clubs addressing: 1) when and where it
is appropriate to charge for football match policing, 2) how much can be charged and
3) how decisions regarding the level of policing services needed are rendered. The
relevant Chief Constable handles these issues for each club through a formalised
annual Charging Agreement that is signed by a representative of the club prior to the
start of a season. This document includes:
1) A request by the football club for special police services under S25.
2) The fixture list of matches for the season.
3) A map of the ground and locations around the stadium the club are
responsible for paying for is clearly identified.
4) A schedule of planned deployments for each category of match for the ground
with standard charges for the categories.
5) A schedule of chargeable rates for the season.
6) A statement to include variations to match categories and additional matches.
7) A set of key contacts for both police and the club who are authorised to deal
with urgent items.
The first underlines each club’s responsibility to satisfy the Chief Constable’s
assessment of ‘proper and adequate’ levels of policing as established by The Guide to
Safety at Sports Grounds 5 Edition 2008. Special policing services are a necessary
prerequisite for obtaining a safety certificate that allows a stadium to host football
matches. The match categorisation system establishes and conveys, amongst other
things, the minimum police staffing requirements to the club.
Match categorisation is integral to understanding the Charging Agreement. Each
match is categorised by the perceived risk of crime and disorder as well as the level of
policing deemed necessary by the Chief Constable. The five standard match
categories are: Category CS (Club Security Only /Police Free), Category A (Low
Risk), Category B (Medium Risk), Category C (High Risk) and Category C-IR
(Category C with Increased Risk).
As the assessed risk of crime and disorder increases so too does the match category
and thus the cost incurred by the club. Predetermined charges and the number of
deployed officers are attached to each category.
14
Intelligence and attendance statistics subsequently dictate the categories into which
each match is assigned. The Chief Constable must demonstrate a rational and
systematic approach for assigning any match to a given category. A schedule of all
fixtures along with their corresponding match categorisation ostensibly provides the
club with an explanation of anticipated costs for the entire season. However, while
fixture lists are relatively stable, many factors can alter their dates and times leading
to the re-categorisation of matches which in turn changes policing costs.
The GMP vs. Wigan AFC rulings established that the Chief Constable can no longer
re-categorise a match with the expectation of compensation for the increased costs.
The assumption that, by accepting the safety certificate, the club willingly accepts
both the police categorisation of the match and the associated costs is no longer
sufficient for recovering policing costs. Rather, formal consent from the club is
required prior to the match. If the club does not agree, the police cannot recover any
additional expenditure beyond the original agreement. The Chief Constable then must
decide to either cover extra costs from the police budget or to leave the match
categorisation unchanged (and accept the risks accompanying under policing). This
reality resulted in the variation statement; a formal notice provided to the clubs prior
to the start of a season that match categories are subject to changes that would make
additional police services (for whose payment clubs remain responsible) necessary.
The variation statement attempts to prevent clubs from refusing reasonable and
informed changes that may arise during a season.
There are several circumstances under which changes to the Charging Agreement
might occur. If a given match is selected for television, resulting in a later kick-off
time, it gives match attendees more time to consume alcohol and engage in disorder.
Likewise, club promotion or relegation can change match categorisation. Moreover, if
a club must win a given match to avoid relegation to a lower league, the risk of
disorder (such as a large pitch invasion) increases due to heightened tensions in the
crowd. Such instances would warrant match re-categorisation to a higher risk level.
However, matches can also be downgraded, lowering the associated policing costs.
For example, harsh weather and intelligence suggesting fewer match attendees than
originally anticipated could lead to a lowered categorisation.
Cup competitions are problematic since neither the police nor the club can know at
the beginning of a season how far that team will advance. Although such matches are
not included in the schedule of costs, variation statements establish that if the club
participates in such matches, they are responsible for the costs associated with
policing them.
GMP v Wigan AFC also deals with the spatial deployment of police at football
matches, commonly referred to as the ‘footprint’. It established that the ‘footprint’
typically consists of the private land owned by the club (e.g. the ground and car
parks), land leased by the club as well as any land or highway that the club controls as
part of the match day. Moreover, clubs requesting police services to other parts of the
immediate surroundings would bear the costs of that policing.
15
Appendix Two: Football Match and Comparison Days
Match Day
Comparison Day
Match Day
Comparison Day
09/08/2005
02/08/2005
22/01/2008
15/01/2008
13/08/2005
30/07/2005
02/02/2008
16/02/2008
13/09/2005
06/09/2005
12/02/2008
05/02/2008
17/09/2005
03/09/2005
04/03/2008
26/02/2008
20/09/2005
04/10/2005
22/03/2008
12/04/2008
01/10/2005
08/10/2005
01/04/2008
25/03/2008
18/10/2005
11/10/2005
14/04/2008
07/04/2008
24/10/2005
17/10/2005
04/05/2008
27/04/2008
05/11/2005
12/11/2005
09/08/2008
02/08/2008
09/11/2005
02/11/2005
12/08/2008
05/08/2008
22/11/2005
15/11/2005
23/08/2008
06/09/2008
26/11/2005
24/12/2005
13/09/2008
11/10/2008
28/12/2005
21/12/2005
20/09/2008
18/10/2008
02/01/2006
09/01/2006
30/09/2008
23/09/2008
07/01/2006
28/01/2006
19/10/2008
12/10/2008
14/01/2006
04/02/2006
28/10/2008
04/11/2008
31/01/2006
24/01/2006
29/11/2008
27/12/2008
11/02/2006
04/03/2006
09/12/2008
02/12/2008
01/04/2006
29/04/2006
26/12/2008
19/12/2008
15/04/2006
06/05/2006
31/01/2009
28/02/2009
22/04/2006
13/05/2006
17/02/2009
10/02/2009
08/08/2006
01/08/2006
03/03/2009
24/02/2009
12/08/2006
05/08/2006
07/03/2009
28/03/2009
23/08/2006
16/08/2006
11/04/2009
02/05/2009
27/08/2006
20/08/2006
18/04/2009
09/05/2009
12/09/2006
05/09/2006
03/05/2009
26/04/2009
23/09/2006
07/10/2006
08/08/2009
01/08/2009
31/10/2006
07/11/2006
11/08/2009
04/08/2009
25/11/2006
09/12/2006
22/08/2009
05/09/2009
28/11/2006
21/11/2006
12/09/2009
19/09/2009
01/01/2007
25/12/2006
15/09/2009
08/09/2009
07/01/2007
31/12/2006
26/09/2009
10/10/2009
20/01/2007
27/01/2007
17/10/2009
31/10/2009
03/02/2007
17/02/2007
20/10/2009
13/10/2009
24/02/2007
10/03/2007
07/11/2009
14/11/2009
06/03/2007
27/02/2007
19/12/2009
09/01/2010
13/03/2007
20/03/2007
09/02/2010
02/02/2010
17/03/2007
24/03/2007
16/02/2010
23/02/2010
09/04/2007
02/04/2007
06/03/2010
27/03/2010
21/04/2007
05/05/2007
20/03/2010
17/04/2010
06/05/2007
29/04/2007
24/03/2010
17/03/2010
19/08/2007
12/08/2007
05/04/2010
29/03/2010
28/08/2007
21/08/2007
18/04/2010
11/04/2010
01/09/2007
08/09/2007
02/05/2010
25/04/2010
18/09/2007
11/09/2007
22/09/2007
13/10/2007
26/09/2007
19/09/2007
23/10/2007
16/10/2007
27/10/2007
17/11/2007
10/11/2007
08/12/2007
27/11/2007
20/11/2007
22/12/2007
05/01/2008
01/01/2008
25/12/2007
19/01/2008
26/01/2008
Comparison Day Legend= 7 days; 14 days; 21 days; 28 days
16
Appendix Three: Technical Overview of Statistical Methods
Permutation tests were used to examine whether the overall count of crime events
was higher on match days than on non-match days. In the first instance, for such a test
we compare the overall difference between the total count of events observed across
match days and non-match days with what would be expected if the probability of an
event occurring on either type of day was the same. The aim is to compute the
probability of seeing a difference in crime counts, beyond that which was observed, if
the probability of an offence occurring was independent of whether a match took
place.
To compute the probability, first we took the observed difference for the total count
of crimes between match and non-match days. Next, we summed the total number of
observed crimes that took place on both match and non-match days and obtained a
sample distribution by randomly generating pairs of numbers that when summed
equalled the total number of observed crimes. To increase the reliability of our
estimates, this procedure was performed 1,000 times computing a difference for each
permutation. Finally, to determine statistical significance, we used our sample
distribution to calculate the fraction of the observed differences that had a difference
equal to or greater than the original observed difference between the match and non-
match comparison days.
One assumption of the above approach is that the difference in the total observed
count of crimes for match and non-match comparison is roughly equivalent, and
hence that what is observed in the aggregate will generally apply for each pair of
match-to-non-match day comparisons. This may or may not be a reasonable
assumption. In the extreme, it is possible that for a single match-to-non-match
comparison there may be a huge difference in the counts of events for the match day
and its associated non-match comparison day (e.g. 10 vs 200 events). For this reason,
we used a variation of the above test which was based on different assumptions. Thus,
for each match-to-non-match comparison we calculated the difference in the observed
counts. We then compared this with what would be expected if the observed daily
counts were preserved but the type of day on which they occurred (match versus non-
match) was random.
To compute the probability, first we took the sum of the observed differences for the
count of crimes between match and non-match day pairs. Next, we randomly
reassigned each of the preserved match and non-match day pair counts into either a
pseudo-match or pseudo non-match day sample to generate the distribution for the
observed differences. Again this was performed 1,000 times with the sum of the
observed difference between pseudo-match and pseudo-non-match day pairs
computed for each permutation. As with the first abovementioned permutation test,
we determined statistical significance by calculating the fraction of the observed
pseudo-match and pseudo-non match day differences that had a difference equal to or
greater than the original observed difference between the match and non-match day
pairs.
17
The two approaches tell us different things. The first approach answers the question
“in total were there more crimes on match days than non-match days”. The second
approach concerns the question “in general, is it the case that more crimes
consistently occur on match days than on comparable non-match days”. By using
both approaches we can triangulate the results. Where consistencies are observed we
can therefore be more precise about what the results mean, and more confident in our
conclusions. However, because the results for both types of analysis are generally
consistent, for simplicity in the main body of the text we report only those results for
the second type of test. For completeness, we include both sets of results in this
appendix.
In addition to estimating the probability that the observed difference in crime counts
was due to chance, it is also possible to examine the same questions for subsets of the
data using both of the aforementioned permutation tests. For example, we used the
test to estimate the probability of seeing a difference equal to or greater than that
observed for events that occurred within six specific hourly intervals: 0 to 3, 4 to 7, 8
to 11, 12 to 16, 17 to 20 and 21 to 24. Moreover, we used the test to determine the
probability of seeing a difference equal to or greater than the original observed
difference between the match and non-match days within different spatial buffers.
A Kolmogorov-Smirnov model was used to test if differences existed in the spatial
distributions of crimes between match and non-match days. Unlike the permutation
models described above, the Kolmogorov-Smirnov test examines the distribution of
crimes between samples - the match day events and non-match day events - and
determines if there are statistically significant differences between the two.
Theoretically, it is possible that there is a significant difference in the crime counts
for a given area on match days versus non-match days, but that the spatial distribution
of crime in the same area is not significantly different. That is to say it is possible that
a large enough percentage of events still take place within the same areas that they
typically take place when there is no football match and therefore the distributions
may be similar. Conversely, if there is a large enough difference in where crime
events take place on a match day as compared to where events occur on non-match
days, then we conclude that there are differences in the overall distribution of events.
Appendix Four: Spatial Permutation Significance Result Table
Permutation Test 1
Buffer
0-250
250- 500
500- 750
750-1000
1000-1250
1250-1500
1500-1750
1750-2000
2000-2250
2250-2500
2500-2750
2750-3000
Area (hectares)
20
59
98
137
177
216
255
295
334
373
412
452
Incidents (All)
**
Crimes (All)
***
Violence
**
Theft/handling
***
Criminal damage
***
Other
*
Permutation Test 2
Incidents (All)
**
*
Crimes (All)
***
Violence
**
*
Theft/handling
**
Criminal damage
**
Other
**
*=p<0.10, **=p<0.05, ***=p<0.01, ****=p<0.001, *****=p<0.0001
19
Appendix Five: Temporal Permutation Significance Result Table
Permutation Test 1
Hour Intervals
0 to 3
4 to 7
8 to 11
12 to 15
16 to 19
20 to 23
Incidents (All), 500m
**
*
***
Crimes (All), 1500m
***
***
***
Violence, 500m
Theft/handling, 750m
*
Criminal damage, none
Other, 1000m
Permutation Test 2
Hour Intervals
0 to 3
4 to 7
8 to 11
12 to 15
16 to 19
20 to 23
Incidents (All), 500m
*
**
**
**
*
Crimes (All), 1500m
***
***
***
Violence, 500m
**
**
Theft/handling, 750m
**
**
Criminal damage, none
Other, 1000m
**
**
*
*=p<0.10, **=p<0.05, ***=p<0.01, ****=p<0.001, *****=p<0.0001
Appendix Six: Detailed Report of Site Visit
An intelligence briefing for the Saturday 16 April 2011 Sheffield Wednesday v
Hartlepool match was attended prior to the 15:00 kick-off. The briefing was attended
by only a few officers because the match category had been downgraded at the
request of Sheffield Wednesday to a CS. This match category meant that the match
would be almost entirely without police. The briefing covered: the number and names
of away coaches. Intelligence suggested there were no Hartlepool ‘risk supporters’
planning to attend the match. A discussion of previous disorder between the two sets
of supporters that resulted in the hospitalisation of a Wednesday supporter was
mentioned but intelligence did not suggest there would be any retaliation. The match
commander explained that a group of additional officers would be taking up positions
throughout the city centre but would be made available at the ground if necessary.
A very small group of officers who were policing the fixture gathered inside
Hillsborough and a pre-match briefing took place. The officers were instructed of
their positions prior to, during and after the match and were informed that if trouble
arose a group of officers patrolling the city centre were ready to be deployed. There
were no questions and the briefing concluded with officers making their way to
various locations around Hillsborough.
Local public houses that Wednesday ‘risk supporters’ frequent were observed. Police
walked through the ‘Travellers’ public house where a group of the youth ‘risk
supporters’ have been seen trying to buy drinks before and after matches throughout
the season. There was relatively little happening at the ‘Travellers’ however two
Wednesday youth ‘risk supporters’ who were underage were identified.
Approximately a half hour after being identified inside the ‘Travellers’ these two
youth ‘risk supporters’ were pulled aside as they attempted to gain entry into
Hillsborough. One of these ‘risk supporters’ was seventeen and informed the officer
that he was trying to break away from the hooligan lifestyle but needed support. The
officer seized this opportunity to introduce a youth worker to the ‘risk supporter’ and
details were exchanged. The youth worker was part of a new early intervention
program that attempts to provide services to youth ‘risk supporters’ from both
Sheffield clubs.
There was virtually no police presence outside the stadium prior to the match. The
supporters queuing up at turnstiles were well behaved and presented no concern for
those police who were present. This may have been the result of a meaningless end of
the season match for both sides who had nothing to play for in terms of promotion or
relegation. According to the Bronze commander in charge the club used this as part of
their argument for downgrading the match category. The police felt that the club’s
argument was compelling and so agreed to the change.
There were no problems prior to, during or after the match. During the match a
discussion took place between the head of club security and police in the control
room regarding the possibility of downgrading the upcoming Wednesday fixture. The
21
Bronze commander indicated that obviously he could not make the decision
unilaterally but would start the necessary process to see if the match could potentially
be re-categorised to a lower risk.
Several officers who were informally interviewed indicated that there was
undoubtedly a different standard for the acceptance of behaviour on football match
days as opposed to other days. When questioned regarding differences between the
‘risk supporters’ of the two Sheffield clubs the officers felt that there was little
difference in behaviour. They did however indicate presently there appeared to be a
larger number of youth ‘risk supporters’ associated with United.
22
REFERENCES
Klinger, David A. 1997. Negotiating Order in Patrol Work: An Ecological Theory of
Police Response to Deviance. Criminology 35 (2):277-306.
Kurland, Justin, Shane Johnson, and Nick Tilley. 2010. Late weekend kick-offs at
premier football matches and their possible link to increased levels of alcohol-
related ASB and disorder. edited by U. D. o. S. a. C. Science. London:
Government Office for London.
Simmons, Jon, Clarissa Legg, and Rachel Hosking. 2003. National Crime Recording
Standard (NCRS): An analysis of the impact on recorded crime Companion
Volume to Crime in England and Wales 2002/2003. In Home Office On-Line
Report 32/03. London: Home Office.
