I'd be interested in knowing whether till-data from pubs, nightclubs, and dance venues followed the same time pattern as they're suggesting in the survey.
In Casswell's story, everybody starts drinking hard at 2 a.m.: the survey shows heavier drinking among those staying at the pub after 2 a.m. because people start drinking more heavily after 2 a.m.. In that world, earlier closing times might reduce harms from heavy consumption. We would then expect to see, within a bar tab, increasing spend after 2 a.m.. I think you'd need to look at patterns within a tab to make a clean call on this: if we only saw increased per patron expenditures, that's equally consistent with Casswell's story (nothing good happens after2 a.m.) and mine (those who are left in the bar after 2 a.m. were drinking more heavily before 2 a.m. as well, and the lightweights have gone home).
I think it's rather more likely that folks who'd intended on going hard started going hard earlier in the evening and then failed to go home before 2 a.m. - the time series is just picking up cohort effects. If my story's right, then we predict that lighter-drinking tabs are more likely to be closed earlier in the evening, that there isn't any particular increase in within-tab expenditures after 2 a.m., but that per patron consumption increases after 2 a.m. because the patron composition changes over the course of the evening. In that world, closing at 2 a.m. doesn't do tons of good because the heavy drinkers already did most of their heavy drinking; they just otherwise would have stuck around the bar for longer. And, a blanket 2 a.m. closing time would do a lot of harm to all-night dance venues where patrons might not be all that likely to be doing much heavy drinking, other than water, after 2 a.m. anyway (MDMA *cough*).
It's at least testable.
In unrelated news, I look forward to tonight's beer tasting at the University Staff Club.