Should the United Kingdom remain a member of the European Union or leave the European Union? (TNS, online)
Remain 40% (-1)
Leave 47% (+4)
The fieldwork for TNS polls is often well out of date by the time of publication, but in this case it concluded yesterday - although it ran for about a week, so it still has older data than the most recent polls from other firms.
In a probably needless attempt to cover themselves, TNS have issued a statement identifying several different reasons why the apparent Leave surge and overall Leave lead may be misleading. Most obviously, they point out the danger that normal sampling variation could give the illusory appearance of a swing towards one side or another - but in reality, we have several polls from different firms conducted at roughly the same time, and all but one of them have shown some kind of movement towards Leave, so it's unlikely that's happening by random chance. TNS also note that their turnout model (which boosts the Leave lead) is based on past voting patterns at general elections, and it's possible that different types of people may be more likely to drag themselves to the polling station in an EU referendum. That's true, but I do think it's slightly odd that they've used young people as an example. Intuitively, it doesn't seem likely to me that young people will be any more enthused about an EU referendum than about a general election. Affluent and highly-educated voters are much more likely candidates to save Remain's bacon - but their tendency to vote in greater numbers is presumably already factored into the turnout model.
This is a very rare example of a poll where Leave have a slight lead among the Scottish subsample (44% to 43%). If that was the actual result, it would of course scupper any strategy for an early repeat of the independence referendum. But the finding should be taken with a pinch of salt - Scottish subsamples are very small, not correctly weighted, and therefore potentially wildly unreliable.