Making licensed venues safer for patrons: what environmental

(T&F) CDAR230104
Drug and Alcohol Review (March 2004), 23, 19 – 29
SPECIAL SECTION: PREVENTION
Making licensed venues safer for patrons: what environmental factors
should be the focus of interventions?
ROSS HOMEL1, RUSSELL CARVOLTH2, MARGE HAURITZ3 GILLIAN MCILWAIN1 &
ROSIE TEAGUE4
1
School of Criminology and Criminal Justice and Key Centre for Ethics, Law, Justice and Governance, Griffith University,
Policy and Projects, Alcohol, Tobacco and Other Drug Services, Queensland Health, 3Mt Gravatt East and 4School of
Applied Psychology, Griffith University, Australia
2
Abstract
The Queensland Safety Action Projects operationalized a problem-focused responsive regulatory model in order to make
nightclubs and other venues safer. A problem-focused approach requires a careful analysis of the total environment of licensed
venues, including drinking and its controls but also the social and physical environments, patron mix and management practices.
We present new analyses of observational data collected in 1994 and 1996 in the north Queensland cities of Cairns, Townsville
and Mackay. Major reductions in aggression and violence were observed, as well as improvements in many aspects of the venue
environment and management practices. We do not argue in this paper that the interventions caused the environmental and
management changes, although we believe this to be true. Rather, our assumption is that whatever caused them, some of the
environmental and management changes were critical to the reductions in aggression. Regression techniques were used to identify
those factors that best explained the declines in aggression. For reduced physical violence four key predictors were identified:
improved comfort, availability of public transport, less overt sexual activity and fewer highly drunk men. For reduced nonphysical aggression, four key predictors were: fewer Pacific Islander patrons, less male swearing, fewer intoxicated patrons
requiring that management be called and more chairs with armrests. The analyses are consistent with the argument that the
control of drinking is necessary but not sufficient to reduce aggression and violence. [Homel R, Carvolth R, Hauritz M,
McIlwain G, Teague R. Making licensed venues safer for patrons: what environmental factors should be the focus of
interventions? Drug Alcohol Rev 2004;23:19 – 29]
Key words: alcohol, licensed venues, nightclubs, regulation violence, safety.
Introduction
The focus of this paper is the regulation of public
drinking places, which provide the settings for between
one quarter and three quarters of all alcohol consumption, depending on the country and the statistical
measures used ([1], Table 3.2). More exactly, our focus
is licensed venues, in particular nightclubs in city centre
entertainment areas in the Australian state of Queensland. We aim to use a reanalysis of the evaluation data
we collected as part of a series of carefully designed
community-based interventions to answer the question:
what were the key observable changes within licensed
environments most closely related to the large declines
in violence and aggression that were recorded between
two waves of observations in 1994 and 1996.
The point of addressing this question is to sharpen
the focus of interventions designed to make licensed
venues safer. The logic of the analyses is that only
environmental and management factors that changed
between 1994 and 1996 were likely to be causally
related to the declines in aggression or violence
Ross Homel PhD, Professor and Head, School of Criminology and Criminal Justice and Deputy Director, Key Centre for Ethics, Law, Justice and
Governance, Griffith University, Australia; Russell Carvolth OAM, BA Hons, Manager, Policy and Projects, Alcohol, Tobacco and Other Drug
Services, Queensland Health; Marge Hauritz PhD, Consultant, Mt Gravatt East; Gillian McIlwain BA (Hons), PhD student, School of
Criminology and Criminal Justice, Griffith University; Rosie Teague BA (Hons), PhD student, School of Applied Psychology, Griffith University,
Australia. Correspondence to Ross Homel, Professor and Head, School of Criminology and Criminal Justice and Deputy Director, Key Centre for
Ethics, Law, Justice and Governance, Griffith University, Australia. E-mail: [email protected]
Received 1 August 2003; accepted for publication 27 October 2003.
ISSN 0959-5236 print/ISSN 1465-3362 online/04/010019–11 # Australian Professional Society on Alcohol and Other Drugs
DOI: 10.1080/09595230410001645529
20
Ross Homel et al.
(whatever other non-observed changes might also have
had a causal impact), and that knowledge of such
factors provides a better basis for prevention planning
than a smorgasbord of risk factors derived from
correlational analyses of cross-sectional data [2].
Whether it was the interventions or other processes
extraneous to the project that caused the extensive
environmental changes does not need to be adjudicated
for the purposes of this paper: it is sufficient to establish
that change in both the environment and in rates of
violence occurred.
Our starting point is not simply that alcohol is a
commodity [1], it is that the settings in which it is
consumed are also part of an intensely competitive
market. Nightclubs and bars do not sell only liquor;
they sell food, entertainment, excitement and perhaps
drugs and they provide—at a cost—spaces for strenuous physical activity, for meeting others, for sexual
liaisons and generally for behaving outrageously with
relative impunity. In short, nightclubs are businesses,
and their business is pleasure. Unfortunately, God
being a Puritan, pleasure has its price. We know well
that drinking is a sin and what the wages are, but the
divine economy does not stop there. Popping pills and
becoming thirsty, competing for women (or men),
being offside with an aggressive bouncer, being
chronically uncomfortable after standing for a long
period, being crushed in a crowd, becoming deafened
by loud music, or growing just simply hungry or tired
can also have negative effects on individual health,
standards of group behaviour and public order. It is not
necessary to be drunk—no one in the place has to be
drunk—to encounter danger on the dance floor or
bullies at the bar.
The point we are making is not simply a light-hearted
observation on Nightclubs We Have Known. Licensed
venues are multi-faceted places that make a great deal
of money for the people that own them. Making them
safer is therefore a challenge with many faces, doubly
difficult if economic interests are at stake. Any
approach to regulating licensed venues must therefore
be based on a realistic analysis of the actual problem.
Putting this another way, we have always been of the
view that the preoccupation of some alcohol researchers
with host responsibility, server training programmes
and so on does not so much miss the point as reflect an
incomplete understanding of the problem. The problem, as we saw it when we began this line of research a
decade ago [3], is the ‘total environment’, which
includes drinking and the other kinds of observable
factors referred to above, but also includes what is not
directly observable. What we originally had in mind
concerning the ‘not directly observable’ was mainly the
hidden deals that are so characteristic of this industry—
in the worst cases, the agreements by police to back off
certain premises if the price is right, but more
commonly the implicit understanding that venues will
be left alone by police and liquor licensing officials if
there is no trouble serious enough to attract media or
political attention. We would now add as part of the
‘not directly observable’ a range of community and
political influences, and the effects of regulatory and
market reform (including national competition policy)
[4,5].
Regulation, if it is to be any use, therefore requires an
understanding of the actual problem in all its aspects,
and the formulation of effective tools for its solution.
This paper focuses on the part of the problem that is
directly observable: the social and physical environment
of nightclubs. The less easily observed aspects, such as
community influences, are discussed in detail elsewhere, in combination with analysis of effective
countermeasures that incorporate community processes and formal regulation (see especially [6 – 8]).
In the next section we describe briefly the Safety
Action Projects in north Queensland and the processes
of data collection, summarising the pre- and postintervention changes in aggression and violence that
were observed in clubs, as well as the numerous
changes that occurred in their physical and social
environments and in management practices. These
data have not been published previously, except in a
limited circulation report [9]. These changes form the
basis for a more detailed statistical investigation of
environmental and management factors that appear to
be linked intimately with the reductions in aggression
and violence. We conclude with some reflections on
critical factors for violence reduction, and on how we
might move closer to the ideal of a problem-focused,
responsive regulatory model.
The North Queensland Safety Action projects
The Safety Action approach was piloted at Surfers
Paradise in an evaluated implementation during 9
months of 1994 [10,11]. The North Queensland
projects were replications of the Surfers Paradise Safety
Action model in Cairns, Townsville and Mackay
between 1994 and 1996 [6]. Aggression and violence
were targeted by addressing the three levels of regulation manipulated in the Safety Action model: supporting improvements in formal state regulation; providing
a mechanism for community support of the formal
regulation; and facilitating self- and peer-regulatory
compliance. This ‘tripartism’ is a feature of a form of
regulation that is responsive to industry context and
structure, regulatory culture and history, and which
also incorporates ‘tit-for-tat’ strategies that combine
punishment and persuasion in an optimum mix, and
‘enforced self-regulation’, in which private sets of rules
written by business (such as Codes of Practice) are
ratified publicly and, when there is a failure of private
Making licensed venues safer for patrons
regulation, are enforced publicly [12]. The interventions focused on: representation to authorities regarding critical regulatory change and refinement; the
building of community level monitoring and informal
processes to improve the quality of operations of
venues; the development of a network of co-operation
and pressure for better practice among formerly
isolated and mistrustful venue operators; and the
enhancement of external safety through improved
attention by authorities to such issues as observability,
lighting and adequacy of public transport.
Data collection
A key evaluation strategy was unobtrusive observation
of aggression and violence, drinking and serving
practices and numerous characteristics of patrons and
of the physical and social environments in licensed
venues pre- and post-intervention. Only the features of
the data collection methods that bear most directly on
the interpretation of the environmental changes are
reported here. These features have to do mainly with
assessing the reliability of the data and whether the data
from the 2 years can be compared validly.
Activities in all the nightclubs and hotels in the
central entertainment areas of Cairns, Townsville and
Mackay were observed by teams of students during
September 1994 (before the interventions) and October
1996 (after the interventions). Cafeterias and restaurants were excluded. All observation sessions were of
about 2 hours’ duration and were unobtrusive. In 1994,
83 visits were made to 28 venues, while in 1996 116
visits were made to 47 venues. More venues were
visited in 1996 than in 1994, especially in Cairns,
because many new venues had been established in that
period, and it was considered important to include all
operating premises in the area. The growth in the
number of venues reflected an influx of tourism
investors and an upsurge in the number of tourist and
entertainment premises generally across the three cities,
particularly in Cairns. Some refurbishment and upgrading was also noted over this period. The original
aim was to visit each establishment three times, once
early in the evening (commencing between 10 p.m. and
midnight), once in the mid-period (commencing
between midnight and 2 a.m.) and once late (commencing after 2 a.m.). This was almost achieved in 1994,
but the greater number of venues in 1996 meant that
the mean number of visits per venue was only 2.5.
Premises visited less than three times were mainly those
that closed early (before 3 a.m., often earlier).
Because aggression and violence tend to occur more
frequently toward closing time, it is important that the
distribution of visits across time periods be equivalent
in 1994 and 1996. This was achieved thus: in 1994 the
distribution was 44.6% early, 34.9% mid-period and
21
20.5% late, while in 1996 the figures were, respectively,
45.7%, 34.0% and 20.3% (w2(2) = 0.08). Nearly all
visits in both years took place on Thursday, Friday and
Saturday nights, with almost identical distributions in
each year.
A structured, systematic observation technique was
employed, based on an observation schedule of 20
pages consisting of hundreds of items. A draft of the
original observation schedule was prepared, based on
the qualitative open coding scheme devised for a 1989
study [3] and on the study by Graham and colleagues
[13]. This was then tested and refined in a series of pilot
visits.
Students recruited from local universities and colleges observed in mixed-sex groups of three or four to
ensure their safety when leaving in the early hours of the
morning. Those recruited tended to be ‘street smart’
and had a general familiarity with licensed venues. At
least three students of Aboriginal or Islander descent
were employed in each city, and these students were
used in venues where non-Indigenous students would
have felt particularly out of place. Students were told in
a general way about the aims of the research, but care
was taken in training in 1996 to avoid setting up an
expectancy that violence levels would be lower than in
1994. Almost none of the observers in 1996 had
performed the 1994 observations, and in any case it
would have been very difficult for observers to have
remembered what levels of violence obtained in 1994.
The use of different observers in the 2 years reduced the
likelihood of conscious or unconscious exaggerations in
post-intervention ratings.
Several training sessions were conducted to ensure
that students were thoroughly familiar with the observation schedule. It was emphasized during training that
observers were there for scientific purposes, and that
although they should act as normal patrons their job was
not to have a good time but to observe as comprehensively and as accurately as possible, and to especially
record details of all incidents of aggression or physical
violence of which they became aware. A limit of one
alcoholic drink per hour was imposed for each observer.
Observers’ responses to items in the observation
schedule were calibrated for consistency within and
across groups. Each observer completed the survey form
in isolation as soon as possible after the visit, and then at
a subsequent meeting inconsistencies between observers were checked and agreement established. Interrater agreement levels exceeded 85%, reflecting the fact
that unreliable items were modified or eliminated in the
original development of the instrument in Sydney in
1991, using large numbers of students [3].
Graham and her colleagues [13] noted especially that
the decision whether a particular incident should be
deemed aggression is one of the major problems of data
collection. In their study, to obtain some consistency
22
Ross Homel et al.
operational guidelines were adopted which stipulated
that an incident would be classified as aggression if it
involved ‘personal violation (verbal insult, unwanted
physical contact), behaviour that was offensive according
to the norms of the place, or a dispute in which the
participants had personal investment’ (p. 281). The same
guidelines were adopted for the present study (and in our
previous research), which means that as in the Vancouver
research there was some variation from establishment to
establishment in the precise operationalization of what
was physical or non-physical aggression.
Observed changes
Aggression and violence (Fig. 1). Very extensive changes
across virtually every domain were observed in the
north Queensland replications, matching or exceeding
the changes observed in Surfers Paradise. Arguments
declined by 28.2%, verbal abuse by 60.4% and
challenges/threats by 40.5% at a time when all these
forms of aggression were increasing in Surfers Paradise
[14]. Trends in physical violence in the north and in
Surfers are shown in Fig. 1 (assault rates declined by
81.2% between 1994 and 1996 in the north).
The physical environment (Table 1). Apart from a general
increase in the total number of venues, the basic
physical infrastructure of venues did not change much
between 1994 and 1996. Thus features such as seating
design, degree of renovation, general appearance and
upkeep, and ‘theme’ did not change. However, many
specific elements of the physical environment, especially those most influenced by day-to-day management
practices, did change, sometimes markedly, and mainly
in the ‘right’ direction. Lighting improved, as did the
spacing and comfort of tables and chairs, ventilation,
the cleanliness of female toilets, and the availability of
taxis and public transport. The biggest changes were to
do with seating style, arrangements and comfort
(improvements in style and spacing but some reductions in comfort), increases in crowding (with larger
numbers of patrons), cleaner female toilets and better
availability of taxis and public transport.
Venue security. The total number of security personnel
did not increase, and police were most conspicuous by
their absence, at least in uniform. This is consistent with
the observations of Homel, Tomsen & Thommeny [3] in
their research in Sydney. The main changes were to do
with bouncer interactions with patrons and the nature of
their patrolling. Generally there was more friendly
interaction, and a trend towards a more cheerful, relaxed
and pleasant demeanour (from 48.3% to 62.1% of visits).
Perhaps the most important change in security arrangements was a trend away from aimless roaming within
venues to a more ‘problem focused’ approach that
directed staff resources to possible points of friction,
such as aisles and bars (from 13.6% to 31.9% of visits).
There was also more stationary patrolling, reducing the
risks of unnecessary confrontations with patrons while
affording the opportunity to keep most of the establishment under surveillance. In keeping with the Codes of
Practice, these strategies were combined with more
rigorous ID checks at the door.
The social environment (Table 2). Comfort levels moved
toward a ‘medium’ rating, with fewer visits being rated
at the extremes of ‘very comfortable’ or ‘uncomfortable’. A possible explanation for this trend is that venue
Figure 1. Observed rates of assault; n = 183 due to missing values.
Making licensed venues safer for patrons
23
Table 1. Statistically significant changes in the physical environment of venues in the three cities combined, 1994 and 1996 (p 5 0.05)
Variable
Seating comfort
Adequate
Too few
Spaced comfortable tables and chairs
Yes
No
Chairs with armrests
Yes
No
Standing room only
Yes
No
Female toilets
Clean
Dirty
Availability of taxis
Available
Limited
None
Availability of public transport
Available
Limited
None
1994 (n = 83) %
or meana
1996 (n = 116) %
or meana
67.6
32.4
52.2
47.8
60.2
39.8
74.1
25.9
37.5
62.5
55.3
44.7
33.3
66.7
56.2
34.8
50.7
49.3
66.7
33.3
48.6
17.6
33.8
71.2
16.2
12.6
1.4
7.2
91.3
21.7
34.0
44.3
pb
Gamma or
effect sizec
0.04
0.31
0.045
0.50
0.04
0.35
0.001
0.58
0.03
0.32
0.001
0.44
0.0001
0.85
a
For some variables, missing values reduce the sample size. Percentages are reported for ordinal variables, and mean values (usually
mean percentages) are presented for numerical variables. The transitions from percentages to means are shown at various points in
the tables.
b
The test of statistical significance is Pearson’s w2 for ordinal data, and the Mann – Whitney test for numerical data.
c
Gamma is presented for ordinal variables, and the effect size (difference between means divided by the pooled standard deviation)
for numerical variables. Both statistics measure the magnitude of the change in the variable.
managers responded to the project by improving
facilities, but these measures were overwhelmed to
some extent by the increase in patronage. Despite the
greater numbers, there appeared to be more ‘wandering
about’ and ‘table-hopping’ by patrons, but without any
increase in bumping or shoving. There were few
changes in entertainment and recreation. Generally
food was more available—an important change in the
context of a move to responsible serving practices—but
it was still only visible in two-thirds of visits in 1996.
Most patrons appeared to be ‘regulars’ or were ‘out
for a big night’, and there was an increase in these
categories in 1996. Individual ‘cheerfulness’ and
‘friendliness’ were also rated as mainly medium or
high in both years for men and women, although it
seems that women especially were enjoying themselves more in 1996. The ‘decorum expectations of
management’ increased markedly, with improvements
particularly apparent in sexual activities and in
negative interchanges between patrons. The most
overt sexual behaviours (such as ‘heavy necking’ and
‘flagrant fondling’—by males or females) became
much less common, but the more flirtatious ‘checking
out’ or ‘chatting up’ activities by men (and to a lesser
extent women) increased in frequency [terms such as
‘heavy necking’ or ‘flagrant fondling by females’ were
coined by the students who developed the original
observation schedule in the Sydney research. They
have stood the test of time as reliable ways of
assessing patron sexual activity in diverse settings,
and are not meant to be pejorative or reflect value
judgements]. Sexual competition also declined, reducing the intensity of one possible factor in aggressive
and violent incidents [2]. Consistent with a less
permissive environment was the trend to lower levels
of rowdiness, swearing and ‘group territoriality’
among both men and women.
Patron characteristics. There were few changes, despite
increases in numbers and in crowding. ID checks
reduced patrons of high school age, but for some reason
Pacific Islanders also reduced in number (from 4.3% to
1.9% of patrons). There were declines in both small
and large crowds, but an increase in patrons on their
own.
24
Ross Homel et al.
Table 2. Selected statistically significant changes in the social environment of venues in the three cities combined, 1994 and 1996 (p 5 0.05)
Variable
Overall comfort
High (very comfortable)
Medium (moderate comfort)
Low (little comfort)
None (uncomfortable)
Bar crowding
High
Medium
Low
None
Table-hopping
Yes
No
Band as entertainment/recreation
Yes
No
Heavy metal music present
Yes
No
Food—small snacks
Yes
No
Food—hot snacks
Yes
No
Decorum expectations of management
High
Moderate
Permissive
Very permissive
No sexual activity by males
Yes
No
Checking out behaviour by males
Yes
No
Heavy necking, touching by males
Yes
No
No sexual activity by females
Yes
No
Heavy necking, touching by females
Yes
No
Flagrant fondling by females
Yes
No
Individual ‘cheerfulness’ of females
High
Medium
Low
None
Social ‘friendliness’ of females
High
Medium
Low
None
1994 (n = 83) %
or meana
1996 (n = 116) %
or meana
25.7
51.4
14.9
8.1
14.8
69.6
14.8
0.9
21.6
28.4
40.5
9.5
21.7
42.6
20.0
15.7
20.7
79.3
34.5
65.5
24.1
75.9
41.4
58.6
4.8
95.2
19.0
81.0
24.1
75.9
51.7
48.3
6.0
94.0
15.5
84.5
6.8
35.6
38.4
19.2
56.5
56.5
22.6
3.5
10.8
89.2
24.1
75.9
55.4
44.6
73.3
26.7
36.1
63.9
18.1
81.9
13.3
86.7
35.3
64.7
34.9
65.1
14.7
85.3
18.1
81.9
5.2
94.8
28.4
55.4
14.9
1.4
48.2
43.0
7.0
1.8
16.2
62.2
18.9
2.7
40.2
44.7
13.2
1.8
pb
Gamma or
effect sizec
0.009
0.03
0.01
0.09
0.04
0.34
0.01
0.38
0.004
0.64
0.00009
0.54
0.04
0.48
0.00004
0.53
0.02
0.45
0.009
0.38
0.004
0.44
0.0005
0.56
0.0008
0.52
0.003
0.60
0.03
0.36
0.007
0.39
(continued overleaf )
Making licensed venues safer for patrons
Table 2.
Variable
Swearing females
High
Medium
Low
None
Swearing males
High
Medium
Low
None
25
(continued )
1994 (n = 83) %
or meana
1996 (n = 116) %
or meana
12.2
16.2
29.7
41.9
3.5
8.8
28.9
58.8
17.6
27.0
27.0
28.4
5.3
17.5
34.2
43.0
pb
Gamma or
effect sizec
0.02
0.38
0.008
0.35
a
For some variables, missing values reduce the sample size. Percentages are reported for ordinal variables, and mean values (usually
mean percentages) are presented for numerical variables. The transitions from percentages to means are shown at various points in
the tables.
b
The test of statistical significance is Pearson’s w2 for ordinal data, and the Mann – Whitney test for numerical data.
c
Gamma is presented for ordinal variables, and the effect size (difference between means divided by the pooled standard deviation)
for numerical variables. Both statistics measure the magnitude of the change in the variable.
Bar staff. Bar staff became much friendlier (‘friendly’
rating: 44.6 – 67.2%; ‘hostile and rude’ rating: 6.8 –
0.9%), they were better matched with patrons in
terms of gender ratios, and were more likely to be
dressed in the house uniform (43.2 – 61.9%). Identification with explicit house policies and the Code of
Practice, symbolized by the use of uniforms, may
have assisted in the move to less permissive practices
with respect to deviant patron behaviour, although
the need to deal with such behaviour apparently
reduced in 1996. However, the need to deal with
aggressive behaviour apparently did not change,
being observed in about 19% of visits in both years,
and nor did the rated abilities of bar staff in this
respect.
Alcohol and drug consumption (Table 3). Neither male
nor female drinking rates appeared to decline, and nor
did female drunkenness decline significantly, but the
incidence of male drunkenness—particularly at ‘high’
levels—reduced sharply. The percentage of visits with
high levels of male drunkenness declined from 40.2%
to 13.8%, while the percentage with low or no male
drunkenness increased from 26.8% to 42.2%. These
results imply that staff intervened in a firm way when
serving men, in order to prevent intoxication. Because
of this intervention, high drinking rates were less likely
to lead to visible intoxication. Neither cover charges nor
the cost of drinks appeared to change. Given the
evidence for moderate increases in patronage between
1994 and 1996, this suggests that the introduction of
the Code of Practices did not make the establishments
any less profitable. Evidence of drug consumption was
noted in about a quarter of all visits, with only a small
reduction between 1994 and 1996.
Host responsibility (Table 3). Methods for dealing with
intoxicated people improved, with intervention by staff
in at least some cases rising from 30.4% to 53.3%.
Techniques that showed statistically significant changes
included delaying service, offering alternatives, denying
service and calling management. Publicity to patrons
improved, with an increase in the use of underage
drinking warnings, Patron Care signs and other forms
of publicity. Promotion of consumption also declined:
topping up or filling empty glasses was not observed at
all in 1996; ‘happy hours’ nearly halved in frequency;
promotion of specific drinks and the use of gimmicks
halved. Interventions with intoxicated patrons included
refusal of service and asking for age identification (both
up from 1.8% to 20.0% of relevant incidents), but
several other possible strategies, such as offering nonalcoholic drinks, food or alternative transport were not
used at all in either year, suggesting that there was still
considerable room for improvement.
New analyses of the north Queensland Safety Action data
Regression techniques were used to identify factors that
best explained the declines in aggression and violence
in north Queensland (Fig. 1). The visit was the unit of
analysis (n = 199, 83 in 1994 and 116 in 1996). Results
from the three sites were similar and were combined.
The focus was on physical violence (observed in eight
visits in 1994 and two in 1996; p = 0.012) and all forms
of non-physical aggression (23 visits in 1994 and 22 in
26
Ross Homel et al.
Table 3. Selected statistically significant changes in alcohol consumption and host responsibility in venues in the three cities combined, 1994
and 1996 (p 5 0.05)
Variable
Drunkenness: males
High
Medium
Low
None
Types of drinks
consumed: males
% Beer
% Light beer
% Soft drinks
% Water
Publicity to clientele
Under-age drinking
warning
Yes
No
Publicity to clientele
House policy notice
Yes
No
Promotion of
consumption
Top up/replace or fill
empty glasses
Yes
No
Promotion of
consumption
Gimmicks
Yes
No
Staff intervention with
highly intoxicated
patrons
In every case
Sometimes
No intervention
No highly intoxicated
patrons
Nature of intervention
Refusal of service
Yes
No
Response to intoxicated
and ordering Patrons
Appropriate service
Yes
Sometimes
No
Service delay
Yes
Sometimes
No
Offer of an alternative
Yes
Sometimes
No
1994 (n = 83) %
or meana
1996 (n = 116) %
or meana
40.2
32.9
18.3
8.5
13.8
44.0
31.0
11.2
mean = 54.9
1.0
0.5
0.4
mean = 47.4
4.8
3.0
2.9
28.9
71.1
51.7
48.3
32.5
67.5
7.2
92.8
27.7
72.3
pb
Gamma or
effect sizec
0.0004
0.38
0.05
0.0007
0.0009
0.0008
0.001
0.30
0.54
0.46
0.39
0.45
0.02
0.37
0.003
1.0
0.02
0.38
0.0001
0.27
0.002
0.86
0.01
0.43
0.0001
0.71
0.02
1.0
18.1
81.9
0
100
14.7
85.3
1.2
19.3
47.0
32.5
5.6
16.7
19.4
58.3
1.8
98.2
20.0
80.0
47.4
17.1
35.5
72.1
8.8
19.1
5.6
2.8
91.7
16.7
22.7
60.6
0.0
0.0
100.0
6.7
3.3
90.0
(continued overleaf )
Making licensed venues safer for patrons
Table 3.
Variable
Service denied
Yes
Sometimes
No
Management called
Yes
Sometimes
No
1994 (n = 83) %
or meana
27
(continued )
1996 (n = 116) %
or meana
1.4
0.0
98.6
7.5
16.4
76.1
2.7
0.0
97.3
4.8
7.9
87.3
pb
Gamma or
effect sizec
0.0002
0.90
0.04
.65
a
For some variables, missing values reduce the sample size. Percentages are reported for ordinal variables, and mean values (usually
mean percentages) are presented for numerical variables. The transitions from percentages to means are shown at various points in
the tables.
b
The test of statistical significance is Pearson’s w2 for ordinal data, and the Mann – Whitney test for numerical data.
c
Gamma is presented for ordinal variables, and the effect size (difference between means divided by the pooled standard deviation)
for numerical variables. Both statistics measure the magnitude of the change in the variable.
1996; p = 0.048). Logistic regressions were used for the
physical violence data and ordinary least squares for
non-physical aggression (total incidents per visit of
arguments, challenges or threats and verbal abuse).
Repeated-measures techniques were not appropriate, as
many venues changed radically through refurbishments
or extensions in the 2-year period, a few closed down
and many others opened up.
The first step was to test, one by one, all environmental factors that changed over time for their capacity
to make the year effect for violence or aggression nonsignificant. Thus each factor that showed a statistically
significant change was fitted as a covariate, followed by
year. The dependent variable was either violence
(1 = one or more incidents; 0 = no incidents) or
aggression (ranging from 0 to 6 incidents per visit). A
large number of tests were conducted in this stage,
especially because many variables had missing values
and various recodings were tried to maintain a sample
size close to 199 (including the creation of nominal
scales with a missing values category). The second step
involved putting the surprisingly small number of
candidates from the first step together in a regression
model (with year), in order to arrive at a minimal subset
of factors that thoroughly explained (in a statistical
sense) the declines in aggression and violence.
For reduced physical violence, four key predictors at
the first stage were identified. These were overall
comfort (p = 0.050 for year, adjusted); availability of
public transport (p = 0.058 for year, adjusted); ‘flagrant
fondling by females’ (p = 0.065 for year, adjusted); and
level of male drunkenness (p = 0.058 for year, adjusted). Interestingly, none of these factors was
individually decisive. However, various combinations
of these four variables clearly ‘washed out’ the year
effect in the second stage. The model with all four
variables is shown in Table 4, although virtually any
subset of at least two variables would suffice to ‘explain’
the year effect. It is also necessary to keep in mind that
with only 10 visits in which physical violence was
recorded, fitting a model with six predictors entails a
high degree of statistical instability, so that relatively
little weight should be placed on specific p-values or
regression coefficients.
For reduced aggression, 25 key predictors at the first
stage were identified, many more than for assaults,
reflecting the more frequent occurrence of non-physical
aggression. Forward stepwise regression was used to
identify the subset most strongly related to aggression,
and these were then fitted as covariates, with year
added. Results are shown in Table 5.
How should the ‘calling management with intoxicated and ordering patrons’ be interpreted? Other
analyses suggest that the predictive power of this
variable reflects having to deal with intoxicated patrons
as opposed to having few drunks on the premises, but
none of the measures of actual intoxication worked as
well as calling management. In other words, the
variable is a surrogate for the number of intoxicated
patrons. The lack of chairs with armrests seems to be
serving as an indicator of the importance of comfortable
places to sit, so taps into the general dimension of
comfort. The most critical thing to note is that year
became clearly non-significant when the four patron
and environmental factors were fitted first. As with
assaults, two or more of these factors was sufficient to
statistically explain the year effect. It is also important
to note that, not surprisingly, aggression was a strong
predictor of assaults (r = 0.53), and that if aggression is
fitted as a covariate with assaults as dependent variable,
the p-value for year drops to 0.071. This suggests that
assaults could be reduced by directly focusing on the
28
Ross Homel et al.
predictors in Table 4, or indirectly by focusing on the
predictors in Table 5.
Discussion
Although very many features of the social and physical
environments of licensed venues changed after the
intervention, the regression analyses suggest that a
relatively small number of factors operating across
disparate domains might be critical to reductions in the
rates of aggression and violence. These factors include
the serving and consumption of alcohol, physical
comfort, the degree of overall ‘permissiveness’ in the
establishment, the availability of public transport, and
aspects of ‘the ethnic mix’ of patrons. It is noteworthy
that these factors are similar to those identified as
correlates of aggression in previous studies over a 20year period in Canada and Australia [2,10,11,13]. This
suggests that the findings are not restricted in their
relevance to one place or time period.
The regression analyses are consistent with our
hypothesis that if one concentrated only on the control
of drinking, reductions in aggression and violence
would not be as great as could be achieved if a more
holistic approach were adopted. Key aspects of drinking
patterns seem to be high levels of male drunkenness
and intoxicated patrons causing so much trouble that
calling management is necessary. Thus the analysis
helps to sharpen the focus of programmes designed to
modify alcohol consumption: whatever controls are
desired on overall levels of consumption, it is essential
that the number of very drunk and troublesome males
be kept to a minimum.
The civilizing impact of physical comfort is consistent with the early qualitative research conducted in
Sydney [3]. However, not all indicators of comfort
moved in a consistent direction (see Tables 1 and 2),
with evidence of deterioration in some respects. This
perhaps reflected a conflict between increased patronage and venue comfort. Nevertheless, the importance of
comfortable seating being available was clear, a
conclusion that will come as no surprise to those who
have spent many hours in clubs and hotels.
Patron behaviour, in the form of overt expressions of
sexual activity and high levels of swearing by men, seem
to reflect a generally permissive environment. The link
between permissiveness and violence makes a lot of
intuitive sense, since if management have an ‘anything
goes’ attitude it is not surprising that violence and
aggression occur. It is noteworthy that venues improved
greatly in these respects after the intervention, and that
casual, non-offensive sexual activity increased substantially, together with overall levels of friendliness and
cheerfulness, especially as expressed by women. Venues
Table 4. Logistic regression model of critical environmental factors explaining the year effect for physical violence
B
Standard
error
df
p
7 2.51
7 1.23
7 0.85
7 1.95
1.36
7 0.10
197.40
1.12
1.23
0.55
0.92
1.10
0.49
971.74
2
1
1
1
1
1
1
1
0.054
0.025
0.320
0.429
0.033
0.215
0.839
0.839
Variable
Comfort
Very, moderately comfortable (relative to missing value)a
Little or no comfort (relative to missing value)a
Male drunkenness (1 = high; 2 = medium; 3 = low; 4 = none)
Flagrant female fondling (1 = yes; 2 = no)
Public transport available (1 = yes; 2 = no)
Year (1 = 1994; 2 = 1996)
Constant
a
These variables were dummy coded.
Table 5. Ordinary least squares regression model of critical environmental factors explaining the year effect for non-physical aggression
Variable
Percentage of Pacific Islander patrons (0 – 30%)
Male swearing (1 = high; 2 = medium; 3 = low; 4 = none)
Call management with intoxicated and ordering patrons
(1 = yes; 2 = sometimes; 3 = no/missing)
Chairs with armrests (1 = yes; 2 = no/missing)
Year (1 = 1994; 2 = 1996)
Constant
B
t
Beta
p
4.94
7 0.21
7 0.43
3.72
7 3.15
7 2.34
0.26
7 0.22
7 0.16
0.000
0.002
0.020
0.30
7 0.02
39.59
2.21
7 0.28
0.29
0.15
7 0.02
0.029
0.784
0.775
Making licensed venues safer for patrons
seem to have become more civil and certainly more
female-friendly, with consequent benefits for safety.
Importantly, these processes do not simply reflect the
effects of alcohol or drunkenness, as they had independent effects in explaining reductions in aggression.
The availability of public transport also makes a great
deal of sense, given the number of incidents we have
witnessed in the streets and on taxi ranks involving
patrons trying to find a way home in the early hours of
the morning. In the north Queensland cities the ready
availability of public transport leaped from 1.4% of
visits in 1994 to 21.7% in 1996 although transport
provided by venues themselves did not improve at all,
highlighting the importance of involving local authorities and communities in these interventions.
A puzzling outcome of the regression analyses was
the finding that reducing the percentage of Pacific
Islander patrons was important in explaining the
reduction in non-physical aggression. This issue is
more fully discussed by Homel & Clarke [10] in an
analysis of data from Sydney licensed venues where
exactly the same relationship was observed. The
phenomenon does appear to be real and may reflect
culturally approved behavioural strategies for dealing
with conflict. This hypothesis requires more research.
It would not be wise to use the results of these
analyses as an infallible guide to regulatory practice.
Incidents of aggression and violence are (fortunately)
relatively rare, and the statistical models must therefore
work within a circumscribed predictive space. Moreover, although the analysis of concomitant changes in
indicators of environmental and behavioural processes
is an improvement on analyses of data collected at a
single time, such a strategy still falls short of experimental designs in probative value. Indeed, a sensible
research strategy would be to now devise a series of
experiments based on the findings of this paper,
probing the power of improved comfort, reduced
permissiveness and so on, to effect improvements in
patron safety.
Such experiments would require the active cooperation of licensees and perhaps of the regulatory
authorities, but would be a useful tool for strengthening
the ecological and systems approach that we and others
advocate (e.g. [15]). The trick is to mobilize resources
at the three levels of the responsive regulatory model
while maintaining a sharp focus on the problem as it
actually is, not as we imagine it to be.
Acknowledgements
The research in this report was supported by the
Queensland Department of Health and the Queensland
29
Police Service: National Campaign Against Drug
Abuse Law Enforcement Fund, and by the Criminology Research Council.
References
[1] Babor T, Caetano A, Casswell S et al. No ordinary
commodity: alcohol and public policy. Oxford: Oxford
University Press, 2003.
[2] Graham K, Homel R. Creating safer bars. In: Plant M,
Single E, Stockwell T, eds. Alcohol: minimizing the harm.
London: Free Association Press, 1997:171 – 92.
[3] Homel R, Tomsen S, Thommeny J. Public drinking and
violence: not just an alcohol problem. J Drug Issues
1992;22:679 – 97.
[4] Power M. The audit society: rituals of verification. Oxford:
Oxford University Press, 1997.
[5] Stockwell T, ed. Alcohol misuse and violence: an examination of the appropriateness and efficacy of liquor licensing
laws across Australia. Report no. 5. Canberra: Commonwealth of Australia, 1994.
[6] Hauritz M, Homel R, McIlwain G, Burrows T, Townsley
M. Reducing violence in licensed venues through community safety action projects: the Queensland experience.
Contemp Drug Problems 1998;25:511 – 51.
[7] Homel R. Review of T. Stockwell, ed. An examination of
the appropriateness and efficacy of liquor-licensing laws
across Australia. Canberra: Australian Government Printing
Service. Addiction 1996;91:1231 – 3.
[8] Homel R, McIlwain G, Carvolth R. Creating safer drinking
environments. In: Heather N, Peters TJ, Stockwell T, eds.
International handbook of alcohol dependence and problems. Chichester: John Wiley & Sons, 2001:721 – 40.
[9] Hauritz M, Homel R, Townsley M, Burrows T, McIlwain
G. An evaluation of the Local Government Safety Action
Projects in Cairns, Townsville and Mackay: a report to the
Queensland Department of Health, the Queensland Police
Service and the Criminology Research Council. Brisbane:
Centre for Crime Policy and Public Safety, Griffith
University, 1998.
[10] Homel R, Clark J. The prediction and prevention of
violence in pubs and clubs. Crime Prevention Studies 3.
Monsey, NY: Criminal Justice Press, 1994:1 – 46.
[11] Homel R, Hauritz M, Wortley R, McIlwain G, Carvolth R.
Preventing alcohol-related crime through community action: the Surfers Paradise Safety Action project. In: Homel
R, ed. Policing for Prevention: Reducing Crime, Public
Intoxication, and Injury. Crime Prevention Studies 7.
Monsey, NY: Criminal Justice Press, 1997:35 – 90.
[12] Ayres I, Braithwaite J. Responsive regulation: transcending
the deregulation debate. New York: Oxford University
Press, 1992.
[13] Graham K, LaRocque L, Yetman R, Ross TJ, Guistra E.
Aggression and barroom environments. J Stud Alcohol
1980;41:277 – 92.
[14] Lincoln R, Homel R. Alcohol and youthful rites of passage.
In: Williams P, ed. Alcohol, young persons and violence.
Canberra: Australian Institute of Criminology, 2001:47 –
60.
[15] Holder HD. Alcohol and the Community: a Systems
Approach. Cambridge: Cambridge University Press, 1998.
S LT
D
230104
T YP
Manuscript No.
ES
E
TT
UE
CDAR
Drug and Alcohol Review
Typeset by Elite Typesetting for
Author
IN G
T EC H
NI
Q
www.elitetypesetting.com
Editor
Master
Publisher
QUERIES:
to be answered by AUTHOR
AUTHOR: The following queries have arisen during the editing of your manuscript.
Please answer the queries by marking the requisite corrections at the appropriate positions
in the text.
QUERY DETAILS
QUERY NO.
No queries
QUERY ANSWERED