Increasing powers to reject licences to sell alcohol: Impacts on availability, sales and behavioural outcomes from a novel natural experiment evaluation

Pliakas, Triantafyllos; Egan, Matt; Gibbons, Janice; Ashton, Charlotte; Hart, Jan; Lock, Karen; (2018) Increasing powers to reject licences to sell alcohol: Impacts on availability, sales and behavioural outcomes from a novel natural experiment evaluation. Preventive medicine, 116. pp. 87-93. ISSN 0091-7435 DOI: https://doi.org/10.1016/j.ypmed.2018.09.010 Downloaded from: http://researchonline.lshtm.ac.uk/id/eprint/4649404/ DOI: https://doi.org/10.1016/j.ypmed.2018.09.010


Introduction
Alcohol accounts for 5.1% of the global burden of disease and 5.9% of deaths worldwide. 1 Excessive alcohol consumption is associated with negative economic and social impacts, including increasing crime, anti-social behaviour and alcohol-related hospital admissions. [1][2][3][4][5] Population level approaches to reduce alcohol availability include modifying economic availability (i.e. taxation), temporal availability (i.e. modify premises' trading times) and spatial availability (i.e. reduce alcohol outlet density). 1,[6][7][8][9][10][11] The mechanisms for modifying alcohol availability differ by jurisdiction but many countries implement some form of licensing system -often regulated at a regional or local level, and often differentiating between licences to sell alcohol for on-premise and for off-premise consumption (in the UK, called 'on-licences' and 'off-licences'). 9,[12][13][14] For many of the world's high alcohol consuming populations, including countries in Europe, Australasia and North America, licensing is an important lever for reducing alcohol availability. 1,15 There is evidence that reduced availability leads to reductions in consumption and harm. 2,16,17 Licensing restrictions have increasingly attracted public health interest for preventing alcohol harms at a populations level.
Regulating alcohol availability in urban areas is particularly important because the world is rapidly urbanising. 18 Major cities often include multiple localities that serve as destinations for the Night Time Economy where consumers come from further afield, but also accommodate large residential populations who are adversely affected by crime and health issues linked to high levels of alcohol availability. 19 If the Licensing system can be used to reduce alcohol availability and harms in London, this should be of interest to alcohol regulators in cities facing similar challenges, both in the UK and elsewhere.
In London, powers to control local alcohol supply and consumption are established by the English Licensing Act 2003. 20 Local 'Licensing Authorities' publish a Statement of Licencing Policy (SLP) every 5 years to show how they plan to meet statutory objectives focused on crime prevention, public safety, public nuisance prevention, and child protection. These SLPs allow discretion with respect to how Licensing Authorities pursue these objectives, taking account of local contexts and priorities.

A C C E P T E D M A N U S C R I P T
2 Licensing Authorities also have discretion to identify sub-areas within their local boundaries ( 'Cumulative Impact Zones' (CIZs)) that will receive more intensive policies ('Cumulative Impact Policies' (CIPs)). CIZs can be created where adverse effects of excess alcohol availability can be demonstrated, with the objective to modify spatial and temporal alcohol availability. CIPs place a greater burden of proof on alcohol licence applicants, requiring them to demonstrate why an application does not undermine its licensing objectives. This puts local authorities in a stronger position to reject or modify a licence application in CIZs. 13 Previous research at the local level in England found that the introduction of a CIP did not decrease successful licence applications 21 but qualitative research has found that some implementers use the powers to modify applications to bring them in line with local policy objectives -for example by approving particular types of premise and seeking reduced hours of alcohol trade. 13 A national study focusing on local licensing policies included CIPs and license application rejections as markers of intervention intensity. The study found that more intense local licensing policies were associated with reductions in alcohol-related hospital admissions and violent and sexual crimes, but not anti-social behaviour. [22][23][24] This research was based on analysis of associations between national datasets and more recently time series analysis using synthetic controls. 24 However, these studies were not designed to evaluate how CIPs might have affected the number and types of licenced premises, changes in temporal alcohol availability, and alcohol sales; nor to differentiate between CIZs and non-CIZs when assessing how these changes effect intermediate behavioural outcomes, such as crime and ambulance call-outs.
CIPs are introduced by local authorities within SLPs that summarise all their different alcohol licensing policy objectives, not just those related to cumulative alcohol impacts.
This can lead to two levels of policy change taking place concurrently: one set of changes implemented across the whole area, whilst additional CIP requirements are implemented in sub-areas identified as CIZs. 13 In this paper we present the results of an impact evaluation of a new alcohol licensing policy in an area of London that has one of the highest densities of pubs, bars, clubs and off-licences in the UK and second highest in London. Alcohol consumption has been identified as a major factor behind crime and disorder in the area with consequences to victims, businesses and local communities. The population suffers from high levels of

A C C E P T E D M A N U S C R I P T
3 alcohol-related ill health and premature deaths. 25 The focus of the new SLP was on implementing an extensive CIP across one local government area and adopting new guidelines of earlier closing times for on and off-licence premises. This paper aims to understand how, and to what extent, the new SLP (including CIP; SLP-CIP) affected alcohol licencing decisions, and how this impacted on temporal, economic and spatial alcohol availability and alcohol-related harms (Figures S1 and S2, Appendix 1 and 2).

Design overview
We

Setting and intervention
In January 2013, a new Licencing Policy implemented a SLP-CIP, which introduced 7 CIZs across the area ( Figure S1, Appendix 1). The new policy also adopted a borough-  Intermediate behaviour outcomes included alcohol related ambulance call-outs using London Ambulance Service data), and overall crime and anti-social behaviour rates using Home Office and open source police data. 26, 27 We calculated crime and alcohol related ambulance call-out rates using mid-year population estimates at Local Authority level. We examined the impact of SLP-CIP on economic availability using on-licence retail sales data (from pubs and bars) to examine effects on alcohol sales, units and quantity of alcohol sold. Data were available for participating premises separately for CIZs and non-CIZs (see Appendix 1 for further details).

Analysis
Statistical analysis was based on interrupted time series analysis undertaken separately for CIZs and non-CIZs and for the whole Local Authority area. For continuous outcomes, we used generalized linear models to obtain Prais-Winsten transformed regression estimators, that are corrected for first-order serially correlated residuals. 28 We used poisson regression for count data. We examined the change in level (i.e. the step change

Characteristics of data
A total of 752 alcohol licence applications were made in the Local Authority area between April 2008 and June 2016, of which 697 (92.7%) had data on decision outcomes and typology (on-or off-license) (Appendix 4, Figure S4). Overall, 78.6% (n=395) and 82% (n=159) applications were granted in CIZs and non-CIZs, respectively (  Figure S4).
Applications for premises in non-CIZs tended to be for off-licences (n=120, 61.9% vs n=194, 38.6%) whilst application in CIZs tended to be for on-licences. The licenced opening hours of premises tended to be longer in CIZs compared to non-CIZs, amongst those licences granted during the study period particularly for on-licence applications.
The total number of overall crime, but not anti-social behaviour, and alcohol-related ambulance call out events and on-licence retail sales were much higher in CIZs compared to non-CIZs (Table 1).

Impact of intervention on licence applications
The introduction of SLP-CIP had no immediate impact on the number of alcohol licence applications submitted (Table 2) but it did result in statistically significant decreases in 'licence approval rates' for total alcohol licence applications in both CIZs (-28.6%, 95% A C C E P T E D M A N U S C R I P T 6 CI = -39.7 to -15.4) and non-CIZs (-24.0%, 95% CI = -41.9 to -0.7) ( Table 3). These were mainly driven by decreases in off-licence applications approved (biannual estimates: -41.4%, 95% CI = -53.2 to -26.5 in CIZs; -24.9%, 95% CI = -44.0 to 0.6 in non-CIZs). These changes were not sustained over time, with 'licence approval rates' for all applications increasing again longer term ( Table 2).
The introduction of SLP-CIP led to immediate, but non statistically significant, reductions in alcohol retail premise opening hours in both CIZs and non-CIZs. Longer term impacts on temporal alcohol availability was observed in non-CIZs but not in CIZs.
Specifically, a reduction of 4.8 hours per week in trading hours (95% CI = -6.6 to -2.9 hours) was observed in non-CIZs, mainly driven by an average reduction in opening hours of on-licence applicants of 10.2 per week (95% CI = -16.0 to -4.5 hours) ( Table 2).

Behavioural Outcomes
Total crime and anti-social behaviour declined over the period when data were available. Our analysis indicated a statistically significant immediate drop in overall crime rates following SLP-CIP. There was a greater drop in CIZs (-12.22%, 95% CI = -17.95 to -6.09) compared to non-CIZs (-7.97%, 95% CI = -13.96 to -1.56). These impacts were reversed over the longer term. Similar trends were observed across the Local Authority (Table 3, Appendix 5). Table 3 shows immediate impacts on anti-social behaviour were not statistically significant, and increased over the longer term both in CIZs and non-CIZs and across the Local Authority. We found no statistically significant impacts on alcohol-related ambulance call-out rates (Table 3).

Sales Outcomes
Our analysis of a sample of on-licence retail sales data found no immediate impact of CIP on sales and units of alcohol sold across the whole study area. In the longer term, there was a significant, but very small, reduction in units of alcohol sold (-2,060, 95% CI = -3,033 to -1,087) but not in total sales value (-664, 95% CI = -3,237 to 1,909) in CIZs.
In non-CIZs, we found a small significant reduction in the sales value (-4,394, 95% CI = -

ACCEPTED MANUSCRIPT
A C C E P T E D M A N U S C R I P T 7 6,421 to -2,367) but not in units of alcohol sold (385.2, 95% CI = -758.5 to 1,528.9).
Similar trends to those in non-CIZ were observed across the Local Authority (Table 4).

Secondary analysis
Findings from this analysis are presented in Appendix 7. There was little difference in licensing outcomes, except from a longer term increase in the duration of trading hours for on-licence applications following the introduction of SLP-CIP. However, this finding should be interpreted with caution as baseline trends in CIZs were significantly different to non-CIZs. The behavioural outcomes did provide more evidence of a difference between CIZ and non-CIZ: notably some evidence of immediate and particularly longer term benefits in terms of reducing anti-social behaviour rates in CIZs compared to non-CIZs (Table S7c). Baseline trends were significantly different between CIZs and non-CIZs for the remaining behavioural outcomes and retail sales data and impact estimates should be interpreted with caution. In summary, the results of the evaluation suggest that overall the SLP-CIP was effective across the whole borough but CIZs, that experienced the policy in a more intense way, did not experience subtantially greater impacts on outcome measures.

Discussion
This study used a natural experimental design to evaluate the impact of the introduction of a new alcohol licensing policy that included CIZs in one of London's major night time economy areas. Our evaluation indicates that the new policy has led to a reduction in temporal availability of alcohol, and an immediate reduction in overall crime rates. The policy did not appear to lead to long term reductions in the number of licence applications submitted or rates of licence applications approved, indicating that SLP-CIP has not been a barrier to gaining a licence to sell alcohol in the area. It suggests successful businesses have been able to adapt to comply with the current Licencing Strategy requirements and, in the case of bars, maintain sales-levels in CIZs even though the amount of unit alcohol consumed fell. We found no policy impact on alcohol-related ambulance call out rates, but this was against a background of declining trends. We observed similar trends in intermediate behaviours outcomes when looking at impacts across the whole Local Authority and separately for CIZs and non-CIZs.

A C C E P T E D M A N U S C R I P T 8
There has been relatively little research considering specific impacts of local licensing policies, 16 and less still focused specifically on CIPs. 21 The most relevant UK study did not distinguish between CIZs and non-CIZs within local authorities when identifying areas of licensing 'intensity.' It did find the local authorities that implemented CIPs and rejected at least one alcohol license application over a two year period experienced moderate reductions in alcohol-related hospital admissions and violent and sexual crimes, although there was inconclusive evidence of an effect on anti-social behaviour. 24 Studies suggest areas categorised as having more intense implementation of alcohol licensing policies have greater reductions in alcohol-related hospital admissions and rates of violent crimes, sexual crimes and public order offences. 22,23 In our evaluation, we found no impact on alcohol-related ambulance call-outs. However, we found that Similar interventions and policies to English SLP-CIPs have been described elsewhere.
In the US the conditional use permits were used to qualify or provide limits on permits on a case-by-case basis if alcohol-related community problems where identified. 30 Other

A C C E P T E D M A N U S C R I P T
9 forms of enforcement include distribution of notification letters to licenced premises, 32 but this is different to zoning restrictions. Government monopolies of alcohol sales have been effective in reducing alcohol-related harms, but difficult or impossible to implement in many settings. 6 In the UK, CIPs have been widely used in local government not necessarily to regulate alcohol outlet density but to influence the diversity of the alcohol retail environment. One study suggests the focus of CIP is to change from predominantly alcohol led drinking establishments towards premises where alcohol consumption is regarded as subsidiary to other forms of consumption such as food. 13,33 There are limitations of this study. The number of submitted applications varied monthly throughout the period of data availability. To overcome this, we aggregated figures at three or six months in order to undertake interrupted time series analysis, which led to some loss in statistical power. As a result of the temporal resolution, we analysed trading hours and closing times, and on-and off-licences using biannual data.
In these models, we did not account for seasonal effects. We did not have data for other local alcohol interventions occurring in the study area at the same time and therefore we cannot account for any potential impact on our results. Prior to Islington's CIP, the Police Reform and Social Responsibility Act (2011) gave local authorities some additional licensing powers and made local health authorities 'responsible authorities,' thus enabling them to make representations against licence applications. These changes are reflected in Islington's 2013 Statement of licensing policy, which is evaluated here. 13, 34 We do know that national duty on alcohol changed during the study period: rising from 2007 then peaking and beginning to fall by 2014. Over the same time, there was a corresponding rise and fall in price for alcohol sold at UK off-licences, but on-license alcohol prices rose throughout the study period. 35 We can not rule out the possibility of boundary effects, for example premises moving to neighbouring areas with less restrictive licensing policies, although neighbouring London Boroughs did have SLP-CIP. 23 Although there had been a small pilot CIZ introduced in 2011, we chose to design the analysis to evaluate the introduction of areawide CIZs as part of a comprehensive licensing policy in 2013. The findings from the retail sales data should be interpreted with caution due to the small sample size and selective nature of premises included in the analysis. We were also unable to assess the impacts of SLP-CIP on alcohol consumption. There are no routine national surveys of

A C C E P T E D M A N U S C R I P T
10 alcohol intake. Although retail sales and rates of alcohol related disease (e.g. liver cancer) are often used as proxy measures, we did not attempt to model intake using them.
Strengths of this study include the use of detailed longitudinal licensing data in combination with detailed intermediate health behaviour data. We have undertaken an in-depth analysis, adjusting for autocorrelation and seasonal effects, where applicable.
Presence of autocorrelation was minimal in almost all models, and further adjustments were made when this was not the case using heteroskedasticity-and autocorrelationconsistent variance estimates. 36 We have also used UK grid coordinate references and latitude and longitude to assign each premise and each event/incident to the CIZs and non-CIZs, which reduces the risk of exposure misclassification. However, it is important to note that geomasking was used for the crime data, which may introduce a higher risk of misclassification. 37

Conclusion
This evaluation shows that more intensive licencing policies can influence alcohol consumption and reduce harms in an area that is a major destination point for consumers of alcohol in one of the world's largest cities. One mechanism for impact could be that more intensive policies encourage license applicants to ensure applications are a 'better fit' with licensing objectives by, for example, asking for fewer hours of sale. Such policies can also lead to reductions in some alcohol-related harms such as crime, while having little negative economic impacts on an area, an important consideration for local government facing austerity. However, determinants and impacts of alcohol consumption are complex, and licencing policies can only shape some aspects of the 'alcogenic environment'. Alcohol policies will need to be multi-sectoral to ensure that alcohol licencing and taxation lead to reductions in alcohol harms at a population level, but are coordinated with focused programmes aimed at higher risk groups.

A C C E P T E D M A N U S C R I P T
Highlights  Modifying alcohol availability is widely advocated to reduce alcohol-related harms.
 Impact evaluation of a new alcohol licensing policy in one London Local Authority.
 First study to examine licensing decisions, alcohol-related harms and retail sales.
 We found moderate reductions in crime and no impact on ambulance callouts.
 We found little or no impact on sales volume and sales revenues.