The paper by Rossow and Norström  adds to the accumulating, although not always consistent, evidence that closing time changes affect levels of violence in bar districts . There is a long history of regulating trading hours to control consumption and problems , with now over 50 studies on the impact of hours of sale . I suggest that it is time to begin focusing efforts on an optimal time for closing of licensed premises.
To identify an optimal closing time, we need to first understand why extending closing time increases violence (or conversely, why restricting closing time reduces violence). The most common (and probable) explanation is that with later closing time, the duration of the drinking occasion is extended and this longer period of drinking results in people drinking more and experiencing more problems. The duration of drinking explanation, however, is based on two assumptions: (a) there is a normative time for starting the drinking session (particularly on weekends) and (b) people start drinking at the same time regardless of closing time.
There appears to be relatively little research on when young adults, who are the most frequent bar-goers , start drinking on nights they go to bars. Some research suggests that later closing times may mean that patrons go to bars later in the evening  but this does not necessarily mean they start drinking later, as evidenced by recent research on predrinking [7–9]. For example, one study that approached young adults in licensed premises at around 9pm found that the majority of young bar patrons started drinking before going out , but this study did not measure the time they actually started drinking.
In terms of research on actual starting time, recent unpublished findings from research on young adult drinking in Switzerland suggests that the average time Swiss young people start drinking on weekend nights range from about 7 to 8pm (Labhart, personal communication), with consumption by men on Saturdays (the heaviest drinking night) increasing steadily until 10–11pm and then leveling off . Research from 4 towns/cities in Lancashire, UK, found similar results, although start times varied from before 5pm (13%), 5 to 7pm (27%), 7 to 8pm (34%) and 8pm or later (35%) . Similar research with young adults in Buffalo, New York [see 12], found that the majority of drinking episodes occurred between 6pm and midnight but with considerable proportions starting earlier, especially on weekends (Collins & Vincent, personal communication).
Although these data are limited, they do suggest that there may be a regular time that young people start drinking. However, it remains unknown as to whether these start times are independent of closing times.
It is also pertinent to ask whether there is a natural time to stop drinking, regardless of how long the bars stay open; that is, people may tend to meander home at a certain time, whether bars close at 2am, 4am or never close. However, a recent study in Newcastle, Australia, found a reduction in violence when closing time changed from 5am to 3:30am , suggesting that people may continue to drink and experience problems as long as bars remain open. Thus, if drinking tends to start at a regular start time, but drinking continues as long as bars remain open, late closing times will have a significant impact on overall consumption and related problems.
On the other hand, closing too early can also have risks associated with displacement of drinking to other private and public locations after the bars close. Despite this risk, few studies have examined possible displacement of drinking and drinking problems with earlier closing of licensed premises. In fact, Rossow and Norström used assaults outside the city centre as the comparison for their evaluation. Thus, the risk of ‘backloading’ as post-bar drinking has been called  also needs to be considered in the overall context of alcohol availability, especially when off-premise establishments stop selling alcohol; that is, backloading would be less likely if alcohol is not for sale elsewhere after bars close.
To conclude, in order to assess whether there is an optimal closing time that minimizes the drinking duration, more research is needed on when people start and stop drinking and whether these times vary by closing time. Moreover, to better understand the impact of closing times on drinking behaviour, more in-depth research is needed on consumption patterns throughout the entire drinking occasion, including factors that influence when people start and stop drinking and their pace of drinking. We also need longer term studies of changes in closing time, because there may be an immediate impact of these changes on drinking duration but this impact may attenuate if people adjust the time they start drinking to match the closing time. Additionally, randomized control experiments on altering closing times are still needed. Natural experiments, regardless of the consistency of findings, cannot rule out potential bias from whatever led to the change in closing time.
Finally, although finding an optimal closing time would be expected to reduce violence and other problems associated with drinking in licensed premises, the size of this effect is likely to be moderate at best. Therefore, research needs to expand the focus beyond hours of sale to other strategies for reducing bar-related violence including developing good transportation options, encouraging well-run establishments with responsible serving practices, strategic policing and reducing environmental risks , and perhaps most importantly changing the culture of drinking that accepts alcohol-related violence as ‘normal’.