Do flexible opening hours reduce violence? An assessment of a natural experiment in alcohol policy

Abstract

BackgroundThroughout history, regulatory approaches to prevent alcohol-related problems have restricted the times and places at which alcohol can be sold. In 2005, the English and Welsh Government took a unique and controversial approach to violence prevention by removing restrictions on opening hours for alcohol outlets. Contrary to traditional restrictive controls, the Licensing Act (2003) proposed to reduce violence by removing restraints on regulated opening hours for licensed premises; thus potentially increasing the physical availability of alcohol. In this study, we treated the removal of trading hour restrictions as a natural experiment in alcohol policy. With a quasi-experimental pattern matching approach, we analysed the Act's effect on violence. MethodsThe study took place in the City of Manchester, UK (population 464 200, 33 electoral wards) during 2004–08. Data for violent incidents were obtained from Greater Manchester Police Service (GMP). We obtained addresses and trading hours for more than 600 licensed premises from magistrates courts (before intervention records) and local government licensing authorities (after intervention records). All data were cleaned and merged into one database with a geographical information system (GIS). We used three strategies to examine the Act's effect on violence. First, we used an interrupted autoregressive integrated moving average (ARIMA) time-series design to estimate the Act's immediate and delayed effect on violence across the city. Sensitivity analysis was done to examine effects of confounding events (ie, police enforcement campaigns, the smoking ban, etc). Second, we did a process assessment to examine how changes to exposure (ie, alcohol availability and staggered closing hours) were affected by the Act, and whether expected changes were achieved. Third, we did a dose-response analysis to examine whether changes to area-level exposures were related to changes in area-level outcomes (ward-level recorded violence). Ordinary least squares regression models adjusted for changes to ward-level outlet density, and spatial autocorrelation was examined to monitor potential geographical spill-over effects. FindingsThe results of our analyses identified little evidence that the deregulation of alcohol opening hours affected citywide violence rates. Analysis of total violence showed no evidence of any immediate, temporary, or delayed intervention effects (ω=0·781, 95% CI −0·06 to 0·17). However, analysis of violence at different times of the day showed a significant 36% increase (gradual and permanent) in weekend violence between 0300 h and 0600 h (δ=−0·897, 95% CI −1·02 to −0·77). The process assessment showed much variation in how areas implemented extra trading hours under the new regulations. Most premises extended trading hours (n=405, 67%), some did not change (n=103, 16%), and some restricted trading hours (n=18, 3%). Only one premise acquired a license to sell alcohol 24 h a day. This variation could have affected the extent to which areas staggered closing times, and increased (or decreased) the availability of alcohol. With a dose-response analysis, we examined whether changes to trading times were associated with changes to area-level violence rates. We noted no relation between changes to staggered closing times, alcohol availability, and area-level violence. InterpretationWe identified no evidence to indicate that the Licensing Act (2003) caused changes to rates of violence in the 2 years after its implementation. Our findings offer little support for fears that increased alcohol availability would lead to substantial increases in violence rates. However, explanations for this counterintuitive finding remain somewhat uncertain because of several limitations, including the short timeframe (2 years), insufficient statistical power to detect small area-level effects, possible under-reporting of violent incidents, and unmeasured and unknown confounding. Many of the limitations encountered in this study could have been averted with improved planning and collaboration between policy makers and scientists to ensure that effective monitoring and assessment was prospectively designed. FundingThe Economic and Social Research Council funded this research with a 1+3 doctoral studentship grant.