RESEARCH REPORT
doi:10.1111/add.12951
The effects of liquor licensing restriction on alcoholrelated violence in NSW, 2008–13 Patricia Menéndez1, Fernando Tusell2 & Don Weatherburn1,2 NSW Bureau of Crime Statistics and Research, Department of Justice, Sydney, Australia1 and School of Economics and Business, University of the Basque Country UPV/EHU, Spain2
ABSTRACT
Aim To estimate the effect on assault of a series of legislative reforms that restricted the trading hours and trading conditions of licensed premises in New South Wales (NSW), Australia. Methods We examine the effects of the legislative reforms introduced between July 2008 and January 2012 using time series structural models. These models are used to estimate the underlying long-term dynamics of the time series of police recorded domestic and non-domestic assaults occasioning actual bodily harm (ABH) and assaults occasioning grievous bodily harm (GBH) in NSW between January 1996 and December 2013. The effect of the legislative changes is captured by including terms in the models which reflect a smooth step change in the number of assaults. Results The reforms introduced between July 2008 and January 2012 were associated with a fall in levels of ABH and GBH assaults. The joint effect of all the interventions on ABH lasted until July 2013, accounting for a reduction of 31.27% over that period [parameter estimate 0.38 with 95% confidence interval (CI) = 0.65, –0.10)]. The same set of interventions had a greater effect on GBH assaults; achieving a 39.70% reduction over a shorter period of time July 2008 and July 2012 (parameter estimate 0.51 with 95% CI = 0.69, – 0.33). Conclusion Legislative reforms introduced in New South Wales, Australia between July 2008 and January 2012 to restrict trading hours and trading conditions of licensed alcohol premises appear to have reduced the number of police-recorded assaults of ABH and GBH by 31.27% and 39.70% respectively. Keywords
Alcohol, assault, consumer sentiment index, liquor licensing, structural time–series.
Correspondence to: Don Weatherburn, NSW Bureau of Crime Statistics and Research, Level 1, Henry Dean Building, 20 Lee st, 2000 Sydney, Australia. E-mail: [email protected] Submitted 19 October 2014; initial review completed 5 February 2015; final version accepted 9 April 2015
INTRODUCTION Evidence gathered from a number of countries shows a strong relationship between alcohol and violence [1,2]. There is also growing evidence that reducing the trading hours of licensed premises is an effective way of reducing alcohol-related harm [3,4]. However, as Kypri et al. [5] pointed out in this journal in 2010, the empirical evidence on alcohol policy is still comparatively small. This is especially true of empirical research surrounding the effect of legislative restrictions on the liquor licence trading hours and conditions. The small number of studies in this area [6–10] generally suggest that restrictions on trading hours reduce alcohol-related harm. Most of these studies, however, involve restrictions on trading hours in a particular location or small number of locations. There is little evidence, to date, on the effect of restricting trading hours and conditions across an entire jurisdiction. © 2015 Society for the Study of Addiction
In 2008, the New South Wales (NSW) Government introduced a number of inter-related reforms designed to encourage more responsible service and consumption of alcohol and reduce the frequency of violence on licensed premises [11]. The first major change occurred on 1 July 2008, with the commencement of the new Liquor Act (2007). This Act gave the Secretary of the Office of Liquor, Gaming and Racing (OLGR) new powers to impose conditions on a liquor licence, to restrict or prohibit liquor promotions and to declare ’lockouts’ and ’curfews’. The second major change, on 30 October 2008, involved the imposition of a freeze on new 24-hour liquor licence trading permits by the Secretary (head) of the Department of Liquor, Gaming and Racing. The package of measures introduced at this time also included a special condition applicable to all liquor licences granted after 30 October 2008, providing that liquor must not be sold on licensed premises for a continuous period of 6 hours during each consecutive 24-hour period. Except where otherwise Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
stipulated, the 6-hour restriction was applied between 4.00 a.m. and 10.00 p.m. The third major change, introduced on 1 December 2008, was the commencement of the ’declared premises’ scheme. This scheme was introduced following the publication of a list of the top 100 licensed premises in terms of numbers of assaults recorded by police as having occurred on those premises [12]. The declared premises scheme imposed special conditions on the 48 licensed premises that had the highest number of recorded assaults between July 2007 and June 2008. The special conditions included: 1 A mandatory 2.00 a.m. lockout of patrons (except for registered clubs). 2 No glass containers to be used after midnight.1 3 No ’shots’2 and drink limit restrictions after midnight. 4 Ten-minute alcohol sales ’time out’ every hour after midnight or active distribution of water and/or food. 5 Cessation of alcohol service 30 minutes prior to closing. The liquor licensing reforms enacted in NSW in July, October and December 2008 were far-reaching and attracted widespread publicity. At the time of their introduction, no similar laws existed anywhere else in Australia. Details of the venues placed on the ’declared premises list’ were publicly released by the NSW Office of Liquor, Gaming and Racing. Publication of the list promoted considerable criticism by licensed premises placed on the list [13,14]. It is of considerable interest to determine, therefore, what effect the new laws had in the overall incidence of violence in NSW. This paper reports the results of an investigation into this issue.
1575
operators and regulators to verify that licensed premises staff had been properly trained in the responsible service of alcohol. On 1 January 2012, a new ’three strikes’ disciplinary scheme for licensed premises was introduced. To isolate and quantify the specific effect of the 2008 interventions, we first model the police-recorded assaults over a period extending from January 1996 to July 2011, just prior to the reforms introduced in August 2011. We then construct models covering the period January 1996 to December 2013 and consider the joint effect of all the interventions described in the paper. The second challenge stems from the fact that only approximately 50% of assaults are reported to police [15]. Changes in public willingness to report assault could interfere with our ability to identify the effect of the 2008 reforms on the trend in assaults. To deal with this problem, we focus on two specific forms of assault: domestic and non-domestic assault occasioning actual bodily harm (ABH) and grievous bodily harm (GBH). According to the Judicial Commission of New South Wales, actual bodily harm includes ’any hurt or injury calculated to interfere with the health or comfort of the victim’, whereas grievous bodily harm is defined ’to include any permanent or serious disfiguring of the person, the destruction of a foetus, and any grievous bodily disease’. We focus on assaults occasioning actual or grievous bodily harm because assaults resulting in injury are much more likely to be reported to police than assaults that do not result in bodily harm or injury [15,16]. Data sets
METHOD Site The site for the study is NSW, a state in Australia. NSW is a state covering 800 642 km2 and which has a population of 7439 million people. More than half this population lives in the Sydney metropolitan area, but substantial numbers live in two other major urban centres north and south of Sydney (Newcastle and Wollongong). General approach There are two main challenges in assessing the effects of the 2008 reforms. The first is that additional changes in liquor licensing policy were made in NSW in 2011 and 2012. On 22 August 2011, a new photo competency card replaced existing paper certificates from that date for students completing approved responsible service of alcohol (RSA) courses. The new RSA database recorded details of courses and students and made it easier for licensed venue
The data for the analysis consisted of monthly counts of the number of domestic and non-domestic ABH and GBH incidents recorded by NSW Police between January 1996 and December 2013. Data for the study were sourced from the NSW Police Force Computerized Operational Policing System (COPS), extracts of which are provided routinely to the Bureau. Monthly data for the same period recording the Consumer Sentiment Index were purchased from the Melbourne Institute. The Consumer Sentiment Index reflects consumers’ evaluations of their household financial situation over the past year and the coming year.3 Model Our analysis of the assault time–series data is based on the socalled time series structural models [17,18] which identify the different features of the data under study in an explicit fashion. The basic underlying idea in time series structural
Introduced to deal with the problem of assaults where a glass was used as a weapon, commonly known as a ‘glassing’. A ‘shot’ is a 30 ml glass of spirits. 3 https://melbourneinstitute.com/miaesr/publications/indicators/csi.html 1 2
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
1576
Patricia Menéndez et al.
Figure 1 Example of three smooth step interventions It between July 2008 and July 2013
models is to explain a time series based on a set of different slow varying components. The time series structural models used in this study to model the outcome variable of assault (ABH or GBH) can be written as the sum of four components. These components are the trend, the seasonal component, the intervention variable and the random errors. The trend represents the model’s long-term dynamics, whereas the seasonal component captures the regular or periodic behaviour which is repeated over time. The intervention variable is designed to capture the effect of the new reforms and the error term describes the model disturbances. In addition, any of these model components can vary over time, making this modelling extremely flexible and efficient to detect changes that occur during the study period. As mentioned above, to quantify the effect of the liquor licensing reforms we include intervention variables in the models to capture the effect of the reforms. Because the 2008 reforms were separated by only a few months, it is impossible to examine the effect of each of them separately without encountering problems of multi-collinearity. We therefore analyse them as a group (i.e. by quantifying their joint effect instead). We assume that if the 2008 reforms have an effect at all, it will be progressive rather than abrupt. We therefore model this process via a smooth step intervention. The smooth step intervention is depicted in Fig. 1 for three particular periods. The detailed models employed in this study and how they relate to autoregressive moving average models, autoregressive integrated moving average (ARMA/ARIMA) models [19] and interrupted time series models [20,21] are described in the Supporting information, together with the state space model representation used to fit the models. The state space model representation [22] allows us to estimate the model components via the Kalman Filter [23] and the model parameters via maximum likelihood estimation, computed using the Kalman Filter. The models in this paper were fitted using R version 3.0 [24] and, in particular, the zoo [25] dlm [26] and forecast [27] packages. The data in this study were aggregated at New South Wales state level and hence we cannot account for possible
heterogeneity of licensing effects or investigate possible clustering impacts. Other than the consumer sentiment index variable, there were no other possible covariates available that would help us to explain the dynamics behind the assaults in this study. As noted earlier, two sets of analyses were carried out in this investigation. In the first, we constructed models for both ABH and GBH containing terms which measure the effect of the 2008 interventions between July 2008 and July 2011. In the second, we model the effects of all the interventions between 2008 and 2012 and examine their effect over different periods of time after the last intervention was introduced.
Model diagnostics The negative log-likelihood values were used as a measure of goodness-of-fit. However, for completeness and to account for model complexity Akaike information criterion (AIC) values were also examined [22,28] and used to select the optimal models.4 The normality assumption concerning residuals was checked via the Shapiro–Wilk test (Ho: residuals are normally distributed) [29]. The independence assumption for the residuals was evaluated using the Box–Ljung test based on the first 15 autocorrelations (Ho: residuals are independent) [30].
RESULTS The reported crime time series for both types of assault (ABH and GBH) are displayed in Fig. 2, together with the time of the interventions. Both time–series show a very strong seasonal pattern and a change in variability over time. In our study, the data were log-transformed to stabilize that variability. The nature of the variation in Fig. 2 is easier to understand if we examine the monthly and yearly trends in reported cases of ABH and GBH assault. The box-plots in the top panel of Fig. 3 show the monthly variation in assault. The yearly variation is shown in the lower panel of Fig. 3.
4
According to the AIC criterion, the smaller the AIC value the better the model.
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
1577
Figure 2 NSW incidents of actual bodily harm and grievous bodily harm assaults between January 1996 until December 2013
Figure 3 Monthly and yearly box-plots of actual bodily harm and grievous bodily harm
Looking at the time series monthly, we can observe that the number of assaults is higher in warmer months (January–March and October–December) than in the cooler months (April–September). The yearly box-plots indicate that both ABH and GBH started to decrease towards the mid-2008. The effect of the 2008 interventions between 2008 July and 2011 July (just before the new set of interventions © 2015 Society for the Study of Addiction
began) can be seen in Table 1, where we observe a 14.59% reduction in ABH and a 31.97% reduction in GBH. The intervention effect is marginally significant in the first case and strongly significant in the second. We cannot discount the possibility that the effect of the 2008 interventions continued after July 2011. However, as two new interventions (policy changes) were introduced in August 2011 and in January 2012, it is not possible from Addiction, 110, 1574–1582
1578
Patricia Menéndez et al.
Table 1 Results for the analysis of actual bodily harm and grievous bodily harm between January 1996 and July 2011.
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
ABH July 2008–July 2011
GBH July 2008–July 2011
0.16 0.09 ( 0.34, 0.02) 1.70 0.09 380.46 0.22 0.45 14.59% 760.76
0.39 0.10 ( 0.59, –0.19) 3.78 0.00 235.67 0.32 0.29 31.97% 471.19
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation.
August 2011 onwards to separate the effect of the 2008 interventions from the later interventions in 2011 and 2012. What we can do is model the conjoint effect of all the interventions that occurred between July 2008 and January 2012. This allows us to determine their combined
effect and duration. Table 2 (ABH) and Table 3 (GBH) show the results of the structural time series modelling for ABH and GBH over different periods of time. The column labelled ’July 2008–July 2012’ in each table shows the results of a model fitted to assault data which assumes the conjoint effects last over the period between July 2008 and July 2012. The column labelled ’July 2008–July 2013’ shows the results of a model which assumes the conjoint effects last from July 2008–July 2013. Finally, the column labelled ’July 2008–December 2013’ shows the results of a model which assumes the conjoint effects last between July 2008 and December 2013. Our interest lies in determining which model best fits the data. Inspection of the row labelled as AIC in Table 2 indicates that the model which best fits the data for ABH is that where the joint effects of the interventions started in July 2008 and finished in July 2013. The estimated joint effect is a (significant) 31.27% fall in the incidence of ABH. The estimated trend displayed in Fig. 4 exhibits the long-term dynamics on ABH assaults for the logtransformed data revealing the true nature of the changes on actual bodily harm levels. It is clear from the figure that up to July 2008 there was a general upwards trend which
Table 2 Results for the analysis of actual bodily harm between January 1996 and December 2013. Models
July 2008–July 2012
July 2008–July 2013
July 2008–December 2013
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
0.25 0.11 ( 0.46, –0.03) 2.19 0.03 381.04 0.40 0.43 22.00% 761.92
0.38 0.14 ( 0.65, –0.10) 2.67 0.01 381.77 0.30 0.49 31.27% 763.39
0.40 0.16 ( 0.71, –0.09) 2.44 0.01 381.63 0.25 0.47 32.76% 763.10
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation.
Table 3 Results for the analysis of grievous bodily harm between January 1996 and December 2013. Models
July 2008–July 2012
July 2008–July 2013
July 2008–December 2013
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
0.51 0.09 ( 0.69, –0.33) 5.69 0.00 236.23 0.12 0.32 39.70% 472.31
0.55 0.12 ( 0.78, –0.31) 4.66 0.00 235.82 0.03 0.30 42.27% 471.48
0.61 0.13 ( 0.86, –0.35) 4.79 0.00 236.63 0.02 0.31 45.52% 473.10
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation. © 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
1579
of GBH is steadily increasing. The decrease in GBH assaults which follows (between July 2008 and July 2012), however, is faster and much more pronounced than the reduction on ABH assaults. The figure indicates a rapid decline in assaults between July 2008 and July 2012, after which the GBH assault trend stabilizes. It is apparent that the combined effect of the interventions for GBH reach stability about 1 year earlier than in the ABH case. DISCUSSION
Figure 4 Logged transform series of actual bodily harm together with the sum of the smoother estimates for the level, slope and the smooth step intervention starting in July 2008 and ending in July 2013
changed from having a positive to a negative slope at the time when the first intervention was introduced. Moreover, we can also see a decrease in assault levels leading to a significant reduction on ABH assaults from July 2008 to July 2013. With regard to GBH, by looking at Table 2 we observe that the best model complying with all the model assumptions (normality and independency of the residuals checked via Shapiro and Box–Ljung tests), and with the smallest AIC value, is the first model, where the joint effect of the interventions lasts from July 2008 until July 2012 and involves a (significant) 39.70% fall in the incidence of GBH. The estimated trend for GBH is shown in Fig. 5. As in the case of ABH, prior to the first intervention the trend
Figure 5 Logged transform series of grievous bodily harm together with the sum of the smoother estimates for the level, slope and the smooth step intervention starting in July 2008 and July 2012 © 2015 Society for the Study of Addiction
The evidence presented here suggests that the legislative reforms to liquor licensing introduced in New South Wales in 2008 produced a substantial drop in the incidence of ABH and GBH in NSW. Prior to the reforms, the assault rate in New South Wales had been rising. Following the reforms the number of assaults began to fall. Our modelling suggests that, by July 2011, the reforms had produced a 14.59% fall in ABH and a 31.97% fall in GBH between July 2008 and July 2011. These are very substantial effects but, given the absence of any comparison group, it is appropriate to consider whether other factors may have contributed to the fall in assault. There were no changes to the law concerning assault over the period of the study, but there are four other factors that could have caused or contributed to the fall in assaults. They are: the ’alcopops’ tax, the global financial crisis, a change in the willingness of staff on licensed premises to report assaults to police and possible changes in social attitudes toward alcohol abuse. The ’alcopops’ tax was an Australian excise duty introduced in 2008 on pre-mixed alcoholic beverages that combine a spirit drink with a soft drink or some form of fruit juice. The tax was introduced in response to concern about the impact of these beverages on teenage drinking. Research conducted by Kiseley et al. [31] suggests that the Australian ’alcopops’ tax had no effect on alcohol-related harms, including intentionally caused injury. Similar findings have been reported in Germany by Muller et al. [32] following the imposition of an ’alcopops’ tax in that country. It therefore seems unlikely that the ’alcopops’ tax was responsible for the drop in assault. The global financial crisis had the potential to influence assault rates because it had a substantial effect on consumer confidence [33], which could have reduced the number of people going out to entertainment precincts (where many assaults occur) or reduce alcohol consumption. To test this possibility, models similar to those described here were fitted, including a measure of consumer confidence (the consumer sentiment index, or CSI) to adjust for possible confounding effects related to global financial crisis The results did not improve the fit significantly and therefore CSI was dropped from the models. The public listing of licensed premises with high rates of assault (and which had their trading hours and conditions Addiction, 110, 1574–1582
1580
Patricia Menéndez et al.
restricted) created a strong commercial incentive not to report assaults on licensed premises to police. The possibility of a decline in willingness to report assaults has been addressed in earlier studies by Moffatt et al. [34] and Snowball & Spratley [35]. Moffatt et al. examined the proportion of assaults reported to NSW police by licensed premises staff during six quarters, between January–March 2008 and April–June 2009, and found no significant change in the proportion of assault reports emanating from licensed premises staff, whether they were one of the top 100 licensed premises (in terms of assault) or one of the licensed premises below the top 100 (in terms of assault).5 Snowball & Spratley extended the analysis up to the October– December quarter 2011 and obtained identical results. This counts strongly against the supposition that the fall in assaults is due to a change in the willingness of staff on licensed premises to report assault to police. A final possibility is that the fall in assaults since 2008 is a result of changes in social or cultural attitudes towards alcohol abuse and/or violence. Without questioning the possibility of such changes, there are two points that count against this interpretation of our results. The first is the abruptness of the change. This is not easily seen in plots of the actual assault rates because of the strong seasonality inherent in the data. Inspection of Figs 4 and 5, however, shows that the underlying trend changed fairly abruptly in July 2008, particularly in the case of assault occasioning grievous bodily harm. If cultural or social change were the underlying mechanism producing the fall in assaults, one might have expected a slower change in the trend. The second consideration that militates against a cultural change interpretation of our results is that earlier research [12] shows that much of the initial change in rates of assault was concentrated in licensed premises that were the focus of new restrictions on liquor licensing. One question that arises from the present study is why the fall in rates of assault occasioning grievous bodily harm (31.97%) was much larger than the fall in rates of assault occasioning actual bodily harm (14.59%) between July 2008 and July 2011. It is difficult to provide a complete or definitive explanation for this finding, but part of the explanation may lie in the fact that a higher percentage of assaults involving grievous bodily harm appear to be alcohol-related. When an assault is believed by attending police to be alcohol-related, they have the option of recording this fact when logging the incident. Between 2005 and 2014, NSW police records show that 53% of reported assaults involving actual bodily harm were recorded as ’alcohol-related’, compared with 59% of assaults occasioning grievous bodily harm.6 Another possibility is that the restrictions on alcohol consumption had bigger effects on
those who consume large quantities of alcohol (and who may be more disposed to violence) than on those who consume smaller quantities. Although the findings from the current study are consistent with those of earlier research in suggesting that restricting liquor outlet trading hours is an effective way of reducing the incidence of assault, the study has two significant limitations. The first is that the initiatives undertaken in 2008, 2011 and 2012 occurred so closely together that we cannot separate their effects. We can be confident that the effects of the 2008 reforms lasted until at least July 2011, but beyond that point the effects of the 2008 reforms are difficult to gauge. The inclusion of intervention terms in the models to account for the policy changes in August 2011 and January 2012 improved the fit of the models up to December 2013. These models revealed a reduction in ABH assaults of 31.27% between July 2008 and July 2013, and a reduction in GBH assaults of 39.70% between July 2008 and July 2012. One could read this as evidence that they helped to prolong or enhanced the effects of the 2008 reforms, but there is no way of knowing for certain. A bigger problem is that, while the central thrust of the 2008 reforms was to restrict the trading hours of licensed premises, the reforms themselves attracted widespread publicity. It is possible that this publicity, and the attendant ’naming and shaming’ of licensed premises with large numbers of assaults, prompted them to serve alcohol more responsibly and/or to take other steps to reduce the number of assaults on or near their premises. This could have contributed to the overall reduction in assaults over and above any impact caused by the restriction on trading hours. It is hard to see how one could ever separate the effects of publicity surrounding restrictions on liquor trading hours from the effects of the restrictions themselves. New laws restricting liquor license trading hours are bound to attract publicity and highly likely to have effects beyond the licensed premises whose trading hours are restricted. In summary, therefore, results presented in this paper show a substantial decrease on assaults in NSW between 2008 and 2013. While we cannot rule out the possibility that the decrease in assaults is due to unmeasured factors, the evidence presented here is consistent with earlier studies finding that liquor licensing restrictions are an effective way of reducing alcohol-related violence.
Declaration of interests The material in this manuscript has not been published in whole or in part elsewhere and is not being considered for publication elsewhere. All authors have been personally and actively involved in substantive work leading to the
5
NSW police are obliged to record details of the person reporting a crime to them. 2015 Unpublished data available on request from the NSW Bureau of Crime Statistics and Research.
6
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
report and will hold themselves jointly and individually responsible for its content. No funds, direct or indirect, have been received from the tobacco, alcohol, pharmaceutical or gaming industries, nor have any funds been received from organizations that seek to provide help with or promote recovery from addiction. The authors are under no contractual constraints regarding the publication of this manuscript. Acknowledgements This research has been funded from the general budget of the NSW Bureau of Crime Statistics and Research.
References 1. Edwards G., Anderson P., Babor T., Casswell S., Ferrence R., Giesbrecht N. et al. Alcohol Policy and the Public Good. Oxford/New York: Oxford University Press; 1994, xii:, pp. 226. 2. Babor T., Caetano R., Casswell S., Edwards G., Gresbrecht G., Graham K. et al. Alcohol: No Ordinary Commodity—Research and Public Policy. Oxford: Oxford University Press; 2003. 3. Middleton J. C., Hahn R. A., Kuzara J. L., Elder R., Brewer R., Chattopadhyay S. et al. The task force on community preventative services, effectiveness of policies maintaining or restricting days of alcohol sales on excessive alcohol consumption and related harms. Am J Prev Med 2010; 39: 575–89. 4. Middleton J. C., Hahn R. A., Kuzara J. L., Elder R., Brewer R., Chattopadhyay S. et al. The task force on community preventative services, effectiveness of policies maintaining or restricting days of alcohol sales on excessive alcohol consumption and related harms. Am J Prev Med 2010; 39: 590–604. 5. Kypri K., Jones C., McElduff P., Barker D. Effects of restricting pub closing times on night-time assaults in an Australian city. Addiction 2010; 106: 303–10. 6. Voas R. B., Lange J. E., Johnson M. B. Reducing high-risk drinking by young Americans South of the border: the impact of a partial ban on sales of alcohol. J Stud Alcohol 2002; 63: 286–92. 7. Voas R. B., Romano E., Kelley-Baker T., Tippetts A. S. A partial ban on sales to reduce high-risk drinking South of the border: seven years later. J Stud Alcohol 2006; 67: 746–53. 8. Douglas M. Restriction of the hours of sale of alcohol in a small community: a beneficial impact. Aust NZ J Public Health 1998; 22: 714–19. 9. Duailibi S., Ponicki W., Grube J., Pinsky I., Laranjeira R., Raw M. The effect of restricting opening hours on alcohol-related violence. Am J Public Health 2007; 97: 2276–80. 10. Rossow I., Norstrom T. The impact of changes in bar closing hours on violence. The Norwegian experience from 18 cities. Addiction 2011; 107: 530–7. 11. Roth L. Liquor licensing restriction to address alcohol-related violence in NSW: 2008–2014, NSW Parliament Research Service E-brief. 2014. Available at: http://www.webcitation.org/ 6X8YYBzux (accessed 19 March 2015). 12. Moffatt S., Mason A., Borzycki C., Weatherburn D. Liquor licensing enforcement and assaults on licensed premises, Bureau Brief no. 40. Sydney: NSW Bureau of Crime Statistics and Research; 2009. © 2015 Society for the Study of Addiction
1581
13. Jensen E. Lockouts increasing the risk of violence. Sydney Morning Herald, 13 January 2009. Available at: http:// www.webcitation.org/6X8Znie4V (accessed 19 March 2015). 14. Kontominas B. Bars take umbrage at laws. Sydney Morning Herald, 4 December 2008. Available at: http://www. webcitation.org/6X8ZwmPMK (accessed 19 March 2015). 15. Australian Bureau of Statistics. Crime Victimisation, Australia, 2012–13. Australia: Catalogue 4530.0; 2014. Available at: http://www.webcitation.org/6X8Z8YCTY (accessed 19 March 2015). 16. Tarling R., Morris K. Reporting crime to the police. Br J Criminol 2010; 50: 474–90. 17. Harvey A. C. Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge UK: Cambridge University Press; 1989. 18. Box G. E. P., Tiao G. C. Intervention analysis with applications to economic and environmental problems. J Am Stat Assoc 1975; 70: 70–9. 19. Box G. E., Jenkins G. M., Reinsel G. C. Time Series and Forecasting. Hoboken, New Jersey: Taylor and Francis; 2013. 20. McDowall D., McCleary R., Hay R. A., Meidinger E. E. Interrupted Time Series Analysis. Beverly Hills, CA: Sage; 1980. 21. Brockwell P. J., Davis R. A.,., editors. Introduction to Time Series and Forecasting. New York, USA: Taylor and Francis; 2002. 22. Durbin J., Koopman S. J. Time series analysis by state space methods (no. 38). Oxford, UK: Oxford University Press; 2012. 23. Kalman R. E. A new approach to linear filtering and prediction problems. J Basic Eng 1960; 82: 35–45. 24. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria 2014. Available at: http://www.webcitation.org/ 6X9wqOrwa (accessed 20 March 2015). 25. Zeileis A., Grothendieck G., Ryan J. A., Andrews F., Zeileis M. A. Package ’zoo’; 2014. R package version 1.7-12. Available at: http://cran.r-project.org/web/packages/zoo/ (accessed 15 May 2015). 26. Petris G. An R package for dynamic linear models. J Stat Software 2010; 36: 1–16. 27. Hyndman R. J. Forecasting functions for time series and linear models. R package version 5.6, 2015. Available at: http://github.com/robjhyndman/forecast (accessed 15 May 2015). 28. Commandeur J. J., Koopman S. J. An Introduction to State Space Time Series Analysis. Oxford: Oxford University Press; 2007. 29. Shapiro S. S., Wilk M. B. An analysis of variance test for normality (complete samples). Biometrika 1965; 52: 591–611. 30. Ljung G. M., Box G. E. On a measure of lack of fit in time series models. Biometrika 1978; 65: 297–303. 31. Kiseley S. R., Pais J., White A., Connor J., Quek L., Crilly J. et al. Effect of the increase in ’alcopops’ tax on alcohol-related harms in young people: a controlled interrupted time series analysis. Med J Aust 2011; 195: 690–3. 32. Muller S., Piontek D., Pabst A., Baumeister S., Kraus L. Changes in alcohol consumption and beverage preference among adolescents after the introduction of the alcopops tax in Germany. Addiction 2010; 105: 1205–13. 33. McDonald T., Morling S. The Australian Economy and the Global Downturn. Part 1: Reasons for Resilience. Licensed from the Australian Government Department of the Treasury under Addiction, 110, 1574–1582
1582
Patricia Menéndez et al.
a Creative Commons Attribution 3.0 Australia Licence. Canberra: Australian Government Department of the Treasury; 2011. 34. Moffatt S., Mason A., Borzycki C., Weatherburn D. Liquor licensing enforcement and assaults on licensed premises. Bureau Brief 2009; 40: 1–12. 35. Snowball L., Spratley S. Is the decrease in assaults at licensed premises being driven by changes in staff reporting rates? Bureau Brief 2013; 87: 1–3.
© 2015 Society for the Study of Addiction
Supporting information Additional supporting information may be found in the online version of this article at the publisher’s web-site: Appendix S1 Study models. Appendix S2 Relationship between state space models, ARIMA and interrupted models.
Addiction, 110, 1574–1582
doi:10.1111/add.12951
The effects of liquor licensing restriction on alcoholrelated violence in NSW, 2008–13 Patricia Menéndez1, Fernando Tusell2 & Don Weatherburn1,2 NSW Bureau of Crime Statistics and Research, Department of Justice, Sydney, Australia1 and School of Economics and Business, University of the Basque Country UPV/EHU, Spain2
ABSTRACT
Aim To estimate the effect on assault of a series of legislative reforms that restricted the trading hours and trading conditions of licensed premises in New South Wales (NSW), Australia. Methods We examine the effects of the legislative reforms introduced between July 2008 and January 2012 using time series structural models. These models are used to estimate the underlying long-term dynamics of the time series of police recorded domestic and non-domestic assaults occasioning actual bodily harm (ABH) and assaults occasioning grievous bodily harm (GBH) in NSW between January 1996 and December 2013. The effect of the legislative changes is captured by including terms in the models which reflect a smooth step change in the number of assaults. Results The reforms introduced between July 2008 and January 2012 were associated with a fall in levels of ABH and GBH assaults. The joint effect of all the interventions on ABH lasted until July 2013, accounting for a reduction of 31.27% over that period [parameter estimate 0.38 with 95% confidence interval (CI) = 0.65, –0.10)]. The same set of interventions had a greater effect on GBH assaults; achieving a 39.70% reduction over a shorter period of time July 2008 and July 2012 (parameter estimate 0.51 with 95% CI = 0.69, – 0.33). Conclusion Legislative reforms introduced in New South Wales, Australia between July 2008 and January 2012 to restrict trading hours and trading conditions of licensed alcohol premises appear to have reduced the number of police-recorded assaults of ABH and GBH by 31.27% and 39.70% respectively. Keywords
Alcohol, assault, consumer sentiment index, liquor licensing, structural time–series.
Correspondence to: Don Weatherburn, NSW Bureau of Crime Statistics and Research, Level 1, Henry Dean Building, 20 Lee st, 2000 Sydney, Australia. E-mail: [email protected] Submitted 19 October 2014; initial review completed 5 February 2015; final version accepted 9 April 2015
INTRODUCTION Evidence gathered from a number of countries shows a strong relationship between alcohol and violence [1,2]. There is also growing evidence that reducing the trading hours of licensed premises is an effective way of reducing alcohol-related harm [3,4]. However, as Kypri et al. [5] pointed out in this journal in 2010, the empirical evidence on alcohol policy is still comparatively small. This is especially true of empirical research surrounding the effect of legislative restrictions on the liquor licence trading hours and conditions. The small number of studies in this area [6–10] generally suggest that restrictions on trading hours reduce alcohol-related harm. Most of these studies, however, involve restrictions on trading hours in a particular location or small number of locations. There is little evidence, to date, on the effect of restricting trading hours and conditions across an entire jurisdiction. © 2015 Society for the Study of Addiction
In 2008, the New South Wales (NSW) Government introduced a number of inter-related reforms designed to encourage more responsible service and consumption of alcohol and reduce the frequency of violence on licensed premises [11]. The first major change occurred on 1 July 2008, with the commencement of the new Liquor Act (2007). This Act gave the Secretary of the Office of Liquor, Gaming and Racing (OLGR) new powers to impose conditions on a liquor licence, to restrict or prohibit liquor promotions and to declare ’lockouts’ and ’curfews’. The second major change, on 30 October 2008, involved the imposition of a freeze on new 24-hour liquor licence trading permits by the Secretary (head) of the Department of Liquor, Gaming and Racing. The package of measures introduced at this time also included a special condition applicable to all liquor licences granted after 30 October 2008, providing that liquor must not be sold on licensed premises for a continuous period of 6 hours during each consecutive 24-hour period. Except where otherwise Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
stipulated, the 6-hour restriction was applied between 4.00 a.m. and 10.00 p.m. The third major change, introduced on 1 December 2008, was the commencement of the ’declared premises’ scheme. This scheme was introduced following the publication of a list of the top 100 licensed premises in terms of numbers of assaults recorded by police as having occurred on those premises [12]. The declared premises scheme imposed special conditions on the 48 licensed premises that had the highest number of recorded assaults between July 2007 and June 2008. The special conditions included: 1 A mandatory 2.00 a.m. lockout of patrons (except for registered clubs). 2 No glass containers to be used after midnight.1 3 No ’shots’2 and drink limit restrictions after midnight. 4 Ten-minute alcohol sales ’time out’ every hour after midnight or active distribution of water and/or food. 5 Cessation of alcohol service 30 minutes prior to closing. The liquor licensing reforms enacted in NSW in July, October and December 2008 were far-reaching and attracted widespread publicity. At the time of their introduction, no similar laws existed anywhere else in Australia. Details of the venues placed on the ’declared premises list’ were publicly released by the NSW Office of Liquor, Gaming and Racing. Publication of the list promoted considerable criticism by licensed premises placed on the list [13,14]. It is of considerable interest to determine, therefore, what effect the new laws had in the overall incidence of violence in NSW. This paper reports the results of an investigation into this issue.
1575
operators and regulators to verify that licensed premises staff had been properly trained in the responsible service of alcohol. On 1 January 2012, a new ’three strikes’ disciplinary scheme for licensed premises was introduced. To isolate and quantify the specific effect of the 2008 interventions, we first model the police-recorded assaults over a period extending from January 1996 to July 2011, just prior to the reforms introduced in August 2011. We then construct models covering the period January 1996 to December 2013 and consider the joint effect of all the interventions described in the paper. The second challenge stems from the fact that only approximately 50% of assaults are reported to police [15]. Changes in public willingness to report assault could interfere with our ability to identify the effect of the 2008 reforms on the trend in assaults. To deal with this problem, we focus on two specific forms of assault: domestic and non-domestic assault occasioning actual bodily harm (ABH) and grievous bodily harm (GBH). According to the Judicial Commission of New South Wales, actual bodily harm includes ’any hurt or injury calculated to interfere with the health or comfort of the victim’, whereas grievous bodily harm is defined ’to include any permanent or serious disfiguring of the person, the destruction of a foetus, and any grievous bodily disease’. We focus on assaults occasioning actual or grievous bodily harm because assaults resulting in injury are much more likely to be reported to police than assaults that do not result in bodily harm or injury [15,16]. Data sets
METHOD Site The site for the study is NSW, a state in Australia. NSW is a state covering 800 642 km2 and which has a population of 7439 million people. More than half this population lives in the Sydney metropolitan area, but substantial numbers live in two other major urban centres north and south of Sydney (Newcastle and Wollongong). General approach There are two main challenges in assessing the effects of the 2008 reforms. The first is that additional changes in liquor licensing policy were made in NSW in 2011 and 2012. On 22 August 2011, a new photo competency card replaced existing paper certificates from that date for students completing approved responsible service of alcohol (RSA) courses. The new RSA database recorded details of courses and students and made it easier for licensed venue
The data for the analysis consisted of monthly counts of the number of domestic and non-domestic ABH and GBH incidents recorded by NSW Police between January 1996 and December 2013. Data for the study were sourced from the NSW Police Force Computerized Operational Policing System (COPS), extracts of which are provided routinely to the Bureau. Monthly data for the same period recording the Consumer Sentiment Index were purchased from the Melbourne Institute. The Consumer Sentiment Index reflects consumers’ evaluations of their household financial situation over the past year and the coming year.3 Model Our analysis of the assault time–series data is based on the socalled time series structural models [17,18] which identify the different features of the data under study in an explicit fashion. The basic underlying idea in time series structural
Introduced to deal with the problem of assaults where a glass was used as a weapon, commonly known as a ‘glassing’. A ‘shot’ is a 30 ml glass of spirits. 3 https://melbourneinstitute.com/miaesr/publications/indicators/csi.html 1 2
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
1576
Patricia Menéndez et al.
Figure 1 Example of three smooth step interventions It between July 2008 and July 2013
models is to explain a time series based on a set of different slow varying components. The time series structural models used in this study to model the outcome variable of assault (ABH or GBH) can be written as the sum of four components. These components are the trend, the seasonal component, the intervention variable and the random errors. The trend represents the model’s long-term dynamics, whereas the seasonal component captures the regular or periodic behaviour which is repeated over time. The intervention variable is designed to capture the effect of the new reforms and the error term describes the model disturbances. In addition, any of these model components can vary over time, making this modelling extremely flexible and efficient to detect changes that occur during the study period. As mentioned above, to quantify the effect of the liquor licensing reforms we include intervention variables in the models to capture the effect of the reforms. Because the 2008 reforms were separated by only a few months, it is impossible to examine the effect of each of them separately without encountering problems of multi-collinearity. We therefore analyse them as a group (i.e. by quantifying their joint effect instead). We assume that if the 2008 reforms have an effect at all, it will be progressive rather than abrupt. We therefore model this process via a smooth step intervention. The smooth step intervention is depicted in Fig. 1 for three particular periods. The detailed models employed in this study and how they relate to autoregressive moving average models, autoregressive integrated moving average (ARMA/ARIMA) models [19] and interrupted time series models [20,21] are described in the Supporting information, together with the state space model representation used to fit the models. The state space model representation [22] allows us to estimate the model components via the Kalman Filter [23] and the model parameters via maximum likelihood estimation, computed using the Kalman Filter. The models in this paper were fitted using R version 3.0 [24] and, in particular, the zoo [25] dlm [26] and forecast [27] packages. The data in this study were aggregated at New South Wales state level and hence we cannot account for possible
heterogeneity of licensing effects or investigate possible clustering impacts. Other than the consumer sentiment index variable, there were no other possible covariates available that would help us to explain the dynamics behind the assaults in this study. As noted earlier, two sets of analyses were carried out in this investigation. In the first, we constructed models for both ABH and GBH containing terms which measure the effect of the 2008 interventions between July 2008 and July 2011. In the second, we model the effects of all the interventions between 2008 and 2012 and examine their effect over different periods of time after the last intervention was introduced.
Model diagnostics The negative log-likelihood values were used as a measure of goodness-of-fit. However, for completeness and to account for model complexity Akaike information criterion (AIC) values were also examined [22,28] and used to select the optimal models.4 The normality assumption concerning residuals was checked via the Shapiro–Wilk test (Ho: residuals are normally distributed) [29]. The independence assumption for the residuals was evaluated using the Box–Ljung test based on the first 15 autocorrelations (Ho: residuals are independent) [30].
RESULTS The reported crime time series for both types of assault (ABH and GBH) are displayed in Fig. 2, together with the time of the interventions. Both time–series show a very strong seasonal pattern and a change in variability over time. In our study, the data were log-transformed to stabilize that variability. The nature of the variation in Fig. 2 is easier to understand if we examine the monthly and yearly trends in reported cases of ABH and GBH assault. The box-plots in the top panel of Fig. 3 show the monthly variation in assault. The yearly variation is shown in the lower panel of Fig. 3.
4
According to the AIC criterion, the smaller the AIC value the better the model.
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
1577
Figure 2 NSW incidents of actual bodily harm and grievous bodily harm assaults between January 1996 until December 2013
Figure 3 Monthly and yearly box-plots of actual bodily harm and grievous bodily harm
Looking at the time series monthly, we can observe that the number of assaults is higher in warmer months (January–March and October–December) than in the cooler months (April–September). The yearly box-plots indicate that both ABH and GBH started to decrease towards the mid-2008. The effect of the 2008 interventions between 2008 July and 2011 July (just before the new set of interventions © 2015 Society for the Study of Addiction
began) can be seen in Table 1, where we observe a 14.59% reduction in ABH and a 31.97% reduction in GBH. The intervention effect is marginally significant in the first case and strongly significant in the second. We cannot discount the possibility that the effect of the 2008 interventions continued after July 2011. However, as two new interventions (policy changes) were introduced in August 2011 and in January 2012, it is not possible from Addiction, 110, 1574–1582
1578
Patricia Menéndez et al.
Table 1 Results for the analysis of actual bodily harm and grievous bodily harm between January 1996 and July 2011.
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
ABH July 2008–July 2011
GBH July 2008–July 2011
0.16 0.09 ( 0.34, 0.02) 1.70 0.09 380.46 0.22 0.45 14.59% 760.76
0.39 0.10 ( 0.59, –0.19) 3.78 0.00 235.67 0.32 0.29 31.97% 471.19
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation.
August 2011 onwards to separate the effect of the 2008 interventions from the later interventions in 2011 and 2012. What we can do is model the conjoint effect of all the interventions that occurred between July 2008 and January 2012. This allows us to determine their combined
effect and duration. Table 2 (ABH) and Table 3 (GBH) show the results of the structural time series modelling for ABH and GBH over different periods of time. The column labelled ’July 2008–July 2012’ in each table shows the results of a model fitted to assault data which assumes the conjoint effects last over the period between July 2008 and July 2012. The column labelled ’July 2008–July 2013’ shows the results of a model which assumes the conjoint effects last from July 2008–July 2013. Finally, the column labelled ’July 2008–December 2013’ shows the results of a model which assumes the conjoint effects last between July 2008 and December 2013. Our interest lies in determining which model best fits the data. Inspection of the row labelled as AIC in Table 2 indicates that the model which best fits the data for ABH is that where the joint effects of the interventions started in July 2008 and finished in July 2013. The estimated joint effect is a (significant) 31.27% fall in the incidence of ABH. The estimated trend displayed in Fig. 4 exhibits the long-term dynamics on ABH assaults for the logtransformed data revealing the true nature of the changes on actual bodily harm levels. It is clear from the figure that up to July 2008 there was a general upwards trend which
Table 2 Results for the analysis of actual bodily harm between January 1996 and December 2013. Models
July 2008–July 2012
July 2008–July 2013
July 2008–December 2013
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
0.25 0.11 ( 0.46, –0.03) 2.19 0.03 381.04 0.40 0.43 22.00% 761.92
0.38 0.14 ( 0.65, –0.10) 2.67 0.01 381.77 0.30 0.49 31.27% 763.39
0.40 0.16 ( 0.71, –0.09) 2.44 0.01 381.63 0.25 0.47 32.76% 763.10
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation.
Table 3 Results for the analysis of grievous bodily harm between January 1996 and December 2013. Models
July 2008–July 2012
July 2008–July 2013
July 2008–December 2013
β1 SD CI t-statistic P-value –Log-likelihood Shapiro test Box–Ljung test Crime reduction AIC
0.51 0.09 ( 0.69, –0.33) 5.69 0.00 236.23 0.12 0.32 39.70% 472.31
0.55 0.12 ( 0.78, –0.31) 4.66 0.00 235.82 0.03 0.30 42.27% 471.48
0.61 0.13 ( 0.86, –0.35) 4.79 0.00 236.63 0.02 0.31 45.52% 473.10
AIC = Akaike information criterion; CI = confidence interval; SD = standard deviation. © 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
1579
of GBH is steadily increasing. The decrease in GBH assaults which follows (between July 2008 and July 2012), however, is faster and much more pronounced than the reduction on ABH assaults. The figure indicates a rapid decline in assaults between July 2008 and July 2012, after which the GBH assault trend stabilizes. It is apparent that the combined effect of the interventions for GBH reach stability about 1 year earlier than in the ABH case. DISCUSSION
Figure 4 Logged transform series of actual bodily harm together with the sum of the smoother estimates for the level, slope and the smooth step intervention starting in July 2008 and ending in July 2013
changed from having a positive to a negative slope at the time when the first intervention was introduced. Moreover, we can also see a decrease in assault levels leading to a significant reduction on ABH assaults from July 2008 to July 2013. With regard to GBH, by looking at Table 2 we observe that the best model complying with all the model assumptions (normality and independency of the residuals checked via Shapiro and Box–Ljung tests), and with the smallest AIC value, is the first model, where the joint effect of the interventions lasts from July 2008 until July 2012 and involves a (significant) 39.70% fall in the incidence of GBH. The estimated trend for GBH is shown in Fig. 5. As in the case of ABH, prior to the first intervention the trend
Figure 5 Logged transform series of grievous bodily harm together with the sum of the smoother estimates for the level, slope and the smooth step intervention starting in July 2008 and July 2012 © 2015 Society for the Study of Addiction
The evidence presented here suggests that the legislative reforms to liquor licensing introduced in New South Wales in 2008 produced a substantial drop in the incidence of ABH and GBH in NSW. Prior to the reforms, the assault rate in New South Wales had been rising. Following the reforms the number of assaults began to fall. Our modelling suggests that, by July 2011, the reforms had produced a 14.59% fall in ABH and a 31.97% fall in GBH between July 2008 and July 2011. These are very substantial effects but, given the absence of any comparison group, it is appropriate to consider whether other factors may have contributed to the fall in assault. There were no changes to the law concerning assault over the period of the study, but there are four other factors that could have caused or contributed to the fall in assaults. They are: the ’alcopops’ tax, the global financial crisis, a change in the willingness of staff on licensed premises to report assaults to police and possible changes in social attitudes toward alcohol abuse. The ’alcopops’ tax was an Australian excise duty introduced in 2008 on pre-mixed alcoholic beverages that combine a spirit drink with a soft drink or some form of fruit juice. The tax was introduced in response to concern about the impact of these beverages on teenage drinking. Research conducted by Kiseley et al. [31] suggests that the Australian ’alcopops’ tax had no effect on alcohol-related harms, including intentionally caused injury. Similar findings have been reported in Germany by Muller et al. [32] following the imposition of an ’alcopops’ tax in that country. It therefore seems unlikely that the ’alcopops’ tax was responsible for the drop in assault. The global financial crisis had the potential to influence assault rates because it had a substantial effect on consumer confidence [33], which could have reduced the number of people going out to entertainment precincts (where many assaults occur) or reduce alcohol consumption. To test this possibility, models similar to those described here were fitted, including a measure of consumer confidence (the consumer sentiment index, or CSI) to adjust for possible confounding effects related to global financial crisis The results did not improve the fit significantly and therefore CSI was dropped from the models. The public listing of licensed premises with high rates of assault (and which had their trading hours and conditions Addiction, 110, 1574–1582
1580
Patricia Menéndez et al.
restricted) created a strong commercial incentive not to report assaults on licensed premises to police. The possibility of a decline in willingness to report assaults has been addressed in earlier studies by Moffatt et al. [34] and Snowball & Spratley [35]. Moffatt et al. examined the proportion of assaults reported to NSW police by licensed premises staff during six quarters, between January–March 2008 and April–June 2009, and found no significant change in the proportion of assault reports emanating from licensed premises staff, whether they were one of the top 100 licensed premises (in terms of assault) or one of the licensed premises below the top 100 (in terms of assault).5 Snowball & Spratley extended the analysis up to the October– December quarter 2011 and obtained identical results. This counts strongly against the supposition that the fall in assaults is due to a change in the willingness of staff on licensed premises to report assault to police. A final possibility is that the fall in assaults since 2008 is a result of changes in social or cultural attitudes towards alcohol abuse and/or violence. Without questioning the possibility of such changes, there are two points that count against this interpretation of our results. The first is the abruptness of the change. This is not easily seen in plots of the actual assault rates because of the strong seasonality inherent in the data. Inspection of Figs 4 and 5, however, shows that the underlying trend changed fairly abruptly in July 2008, particularly in the case of assault occasioning grievous bodily harm. If cultural or social change were the underlying mechanism producing the fall in assaults, one might have expected a slower change in the trend. The second consideration that militates against a cultural change interpretation of our results is that earlier research [12] shows that much of the initial change in rates of assault was concentrated in licensed premises that were the focus of new restrictions on liquor licensing. One question that arises from the present study is why the fall in rates of assault occasioning grievous bodily harm (31.97%) was much larger than the fall in rates of assault occasioning actual bodily harm (14.59%) between July 2008 and July 2011. It is difficult to provide a complete or definitive explanation for this finding, but part of the explanation may lie in the fact that a higher percentage of assaults involving grievous bodily harm appear to be alcohol-related. When an assault is believed by attending police to be alcohol-related, they have the option of recording this fact when logging the incident. Between 2005 and 2014, NSW police records show that 53% of reported assaults involving actual bodily harm were recorded as ’alcohol-related’, compared with 59% of assaults occasioning grievous bodily harm.6 Another possibility is that the restrictions on alcohol consumption had bigger effects on
those who consume large quantities of alcohol (and who may be more disposed to violence) than on those who consume smaller quantities. Although the findings from the current study are consistent with those of earlier research in suggesting that restricting liquor outlet trading hours is an effective way of reducing the incidence of assault, the study has two significant limitations. The first is that the initiatives undertaken in 2008, 2011 and 2012 occurred so closely together that we cannot separate their effects. We can be confident that the effects of the 2008 reforms lasted until at least July 2011, but beyond that point the effects of the 2008 reforms are difficult to gauge. The inclusion of intervention terms in the models to account for the policy changes in August 2011 and January 2012 improved the fit of the models up to December 2013. These models revealed a reduction in ABH assaults of 31.27% between July 2008 and July 2013, and a reduction in GBH assaults of 39.70% between July 2008 and July 2012. One could read this as evidence that they helped to prolong or enhanced the effects of the 2008 reforms, but there is no way of knowing for certain. A bigger problem is that, while the central thrust of the 2008 reforms was to restrict the trading hours of licensed premises, the reforms themselves attracted widespread publicity. It is possible that this publicity, and the attendant ’naming and shaming’ of licensed premises with large numbers of assaults, prompted them to serve alcohol more responsibly and/or to take other steps to reduce the number of assaults on or near their premises. This could have contributed to the overall reduction in assaults over and above any impact caused by the restriction on trading hours. It is hard to see how one could ever separate the effects of publicity surrounding restrictions on liquor trading hours from the effects of the restrictions themselves. New laws restricting liquor license trading hours are bound to attract publicity and highly likely to have effects beyond the licensed premises whose trading hours are restricted. In summary, therefore, results presented in this paper show a substantial decrease on assaults in NSW between 2008 and 2013. While we cannot rule out the possibility that the decrease in assaults is due to unmeasured factors, the evidence presented here is consistent with earlier studies finding that liquor licensing restrictions are an effective way of reducing alcohol-related violence.
Declaration of interests The material in this manuscript has not been published in whole or in part elsewhere and is not being considered for publication elsewhere. All authors have been personally and actively involved in substantive work leading to the
5
NSW police are obliged to record details of the person reporting a crime to them. 2015 Unpublished data available on request from the NSW Bureau of Crime Statistics and Research.
6
© 2015 Society for the Study of Addiction
Addiction, 110, 1574–1582
Liquor licensing restrictions and violence
report and will hold themselves jointly and individually responsible for its content. No funds, direct or indirect, have been received from the tobacco, alcohol, pharmaceutical or gaming industries, nor have any funds been received from organizations that seek to provide help with or promote recovery from addiction. The authors are under no contractual constraints regarding the publication of this manuscript. Acknowledgements This research has been funded from the general budget of the NSW Bureau of Crime Statistics and Research.
References 1. Edwards G., Anderson P., Babor T., Casswell S., Ferrence R., Giesbrecht N. et al. Alcohol Policy and the Public Good. Oxford/New York: Oxford University Press; 1994, xii:, pp. 226. 2. Babor T., Caetano R., Casswell S., Edwards G., Gresbrecht G., Graham K. et al. Alcohol: No Ordinary Commodity—Research and Public Policy. Oxford: Oxford University Press; 2003. 3. Middleton J. C., Hahn R. A., Kuzara J. L., Elder R., Brewer R., Chattopadhyay S. et al. The task force on community preventative services, effectiveness of policies maintaining or restricting days of alcohol sales on excessive alcohol consumption and related harms. Am J Prev Med 2010; 39: 575–89. 4. Middleton J. C., Hahn R. A., Kuzara J. L., Elder R., Brewer R., Chattopadhyay S. et al. The task force on community preventative services, effectiveness of policies maintaining or restricting days of alcohol sales on excessive alcohol consumption and related harms. Am J Prev Med 2010; 39: 590–604. 5. Kypri K., Jones C., McElduff P., Barker D. Effects of restricting pub closing times on night-time assaults in an Australian city. Addiction 2010; 106: 303–10. 6. Voas R. B., Lange J. E., Johnson M. B. Reducing high-risk drinking by young Americans South of the border: the impact of a partial ban on sales of alcohol. J Stud Alcohol 2002; 63: 286–92. 7. Voas R. B., Romano E., Kelley-Baker T., Tippetts A. S. A partial ban on sales to reduce high-risk drinking South of the border: seven years later. J Stud Alcohol 2006; 67: 746–53. 8. Douglas M. Restriction of the hours of sale of alcohol in a small community: a beneficial impact. Aust NZ J Public Health 1998; 22: 714–19. 9. Duailibi S., Ponicki W., Grube J., Pinsky I., Laranjeira R., Raw M. The effect of restricting opening hours on alcohol-related violence. Am J Public Health 2007; 97: 2276–80. 10. Rossow I., Norstrom T. The impact of changes in bar closing hours on violence. The Norwegian experience from 18 cities. Addiction 2011; 107: 530–7. 11. Roth L. Liquor licensing restriction to address alcohol-related violence in NSW: 2008–2014, NSW Parliament Research Service E-brief. 2014. Available at: http://www.webcitation.org/ 6X8YYBzux (accessed 19 March 2015). 12. Moffatt S., Mason A., Borzycki C., Weatherburn D. Liquor licensing enforcement and assaults on licensed premises, Bureau Brief no. 40. Sydney: NSW Bureau of Crime Statistics and Research; 2009. © 2015 Society for the Study of Addiction
1581
13. Jensen E. Lockouts increasing the risk of violence. Sydney Morning Herald, 13 January 2009. Available at: http:// www.webcitation.org/6X8Znie4V (accessed 19 March 2015). 14. Kontominas B. Bars take umbrage at laws. Sydney Morning Herald, 4 December 2008. Available at: http://www. webcitation.org/6X8ZwmPMK (accessed 19 March 2015). 15. Australian Bureau of Statistics. Crime Victimisation, Australia, 2012–13. Australia: Catalogue 4530.0; 2014. Available at: http://www.webcitation.org/6X8Z8YCTY (accessed 19 March 2015). 16. Tarling R., Morris K. Reporting crime to the police. Br J Criminol 2010; 50: 474–90. 17. Harvey A. C. Forecasting, Structural Time Series Models and the Kalman Filter. Cambridge UK: Cambridge University Press; 1989. 18. Box G. E. P., Tiao G. C. Intervention analysis with applications to economic and environmental problems. J Am Stat Assoc 1975; 70: 70–9. 19. Box G. E., Jenkins G. M., Reinsel G. C. Time Series and Forecasting. Hoboken, New Jersey: Taylor and Francis; 2013. 20. McDowall D., McCleary R., Hay R. A., Meidinger E. E. Interrupted Time Series Analysis. Beverly Hills, CA: Sage; 1980. 21. Brockwell P. J., Davis R. A.,., editors. Introduction to Time Series and Forecasting. New York, USA: Taylor and Francis; 2002. 22. Durbin J., Koopman S. J. Time series analysis by state space methods (no. 38). Oxford, UK: Oxford University Press; 2012. 23. Kalman R. E. A new approach to linear filtering and prediction problems. J Basic Eng 1960; 82: 35–45. 24. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria 2014. Available at: http://www.webcitation.org/ 6X9wqOrwa (accessed 20 March 2015). 25. Zeileis A., Grothendieck G., Ryan J. A., Andrews F., Zeileis M. A. Package ’zoo’; 2014. R package version 1.7-12. Available at: http://cran.r-project.org/web/packages/zoo/ (accessed 15 May 2015). 26. Petris G. An R package for dynamic linear models. J Stat Software 2010; 36: 1–16. 27. Hyndman R. J. Forecasting functions for time series and linear models. R package version 5.6, 2015. Available at: http://github.com/robjhyndman/forecast (accessed 15 May 2015). 28. Commandeur J. J., Koopman S. J. An Introduction to State Space Time Series Analysis. Oxford: Oxford University Press; 2007. 29. Shapiro S. S., Wilk M. B. An analysis of variance test for normality (complete samples). Biometrika 1965; 52: 591–611. 30. Ljung G. M., Box G. E. On a measure of lack of fit in time series models. Biometrika 1978; 65: 297–303. 31. Kiseley S. R., Pais J., White A., Connor J., Quek L., Crilly J. et al. Effect of the increase in ’alcopops’ tax on alcohol-related harms in young people: a controlled interrupted time series analysis. Med J Aust 2011; 195: 690–3. 32. Muller S., Piontek D., Pabst A., Baumeister S., Kraus L. Changes in alcohol consumption and beverage preference among adolescents after the introduction of the alcopops tax in Germany. Addiction 2010; 105: 1205–13. 33. McDonald T., Morling S. The Australian Economy and the Global Downturn. Part 1: Reasons for Resilience. Licensed from the Australian Government Department of the Treasury under Addiction, 110, 1574–1582
1582
Patricia Menéndez et al.
a Creative Commons Attribution 3.0 Australia Licence. Canberra: Australian Government Department of the Treasury; 2011. 34. Moffatt S., Mason A., Borzycki C., Weatherburn D. Liquor licensing enforcement and assaults on licensed premises. Bureau Brief 2009; 40: 1–12. 35. Snowball L., Spratley S. Is the decrease in assaults at licensed premises being driven by changes in staff reporting rates? Bureau Brief 2013; 87: 1–3.
© 2015 Society for the Study of Addiction
Supporting information Additional supporting information may be found in the online version of this article at the publisher’s web-site: Appendix S1 Study models. Appendix S2 Relationship between state space models, ARIMA and interrupted models.
Addiction, 110, 1574–1582
